--- title: "Convert to Markdown" description: "Convert to Markdown is the markitdown.ai single-file converter that turns PDFs, Office files, images, HTML, and URLs into structured Markdown." url: https://markitdown.ai/convert updated: 2026-09-26 source: markitdown.ai --- # Convert to Markdown > Convert to Markdown is the markitdown.ai single-file converter that turns PDFs, Office files, images, HTML, and URLs into structured Markdown. The anonymous trial accepts files up to 10 MB and 25 pages, with 3 conversions per IP per hour and 10 a day. Lite starts at $20 per month with files up to 100 MB. Pro is $50 per month; Max is $100 per month. - Structure-aware output for LLMs, RAG, and agents - Preview the rendered Markdown before you copy or download - Server-side conversion pipeline — not fragile browser copy-paste ## Structure that survives conversion The goal is not just to extract text once, but to produce Markdown that downstream tools — ChatGPT, Claude, RAG chunking, and automation — can rely on. - Headings and hierarchy — Section structure stays intact, so the document reads as an outline instead of one undifferentiated block of text. - Tables and lists — Rows, columns, and bullet structure are preserved as Markdown rather than flattened into loose prose. - Reading order — Multi-column pages, headers, and footnotes are resolved into the order a person actually reads. - Preview and source mode — Review the rendered Markdown, then switch to source when you need clean input for a pipeline. ## Supported formats Convert accepts the file types below and always returns Markdown. Every format also has its own page, such as [PDF to Markdown](/pdf-to-markdown). The free public trial takes files up to 10 MB; plan limits are on the [pricing page](/pricing). From Markdown you can go on to [HTML](/markdown-to-html), [plain text](/markdown-to-text), or a [printable PDF](/markdown-to-pdf). - PDF — .pdf — Reports, papers, manuals, and contracts; scanned pages use OCR. - Word — .doc .docx .odt .rtf — Headings, lists, tables, and review comments carry over. - PowerPoint — .ppt .pptx .odp — Slide titles, bullets, and speaker notes, slide by slide. - Excel — .xls .xlsx .ods .csv — Each sheet becomes a Markdown table labelled with its name. - EPUB — .epub — Ebooks, with parts and chapters kept as nested headings. - Images — .png .jpg .jpeg .webp .gif .bmp .tif .tiff — Screenshots, photos, and scans, read with OCR. - HTML — .html .htm — Saved web pages, with headings, links, and tables kept. - Text — .txt .md — Plain text and existing Markdown, normalized. - Web pages — Public URL — Paste a public URL; the page is fetched and converted. ## How conversion works One file, three steps. Longer documents are processed asynchronously, so a large PDF does not block the page. 1. Upload a file or paste a URL — Drop a document in the browser, or point the converter at a public web page. 2. The pipeline rebuilds structure — Conversion runs server-side, so large or complex documents are handled the same way every time. 3. Preview, copy, download, or save — Read the result first, then copy it, download the Markdown, or save it to your library. ## Converted Markdown vs copy-paste Copy-paste from a PDF interleaves columns, flattens tables, and drops headings; ad-hoc extraction leaves page numbers and broken hyphenation. A conversion pipeline reconstructs reading order and structure so the Markdown stays consistent across thousands of documents instead of failing on the messy ones. Need conversion in your own code or in bulk? Use the [developer API](/developers) or [Batch Convert](/batch-convert). ## Traceable output Keep a source filename and revision reference beside the Markdown you reuse. When the source changes, convert the new version and compare the relevant passages instead of silently replacing a reviewed output. This matters when several people use the text or when a retrieved answer needs to cite its origin. A traceable conversion lets a reader check an uncertain table or sentence without guessing which source file produced it. ## Choosing the source Convert to Markdown from the most informative source you have. If the original document is an editable Word file, use that file rather than printing it to PDF first. If a spreadsheet is available, convert the workbook instead of taking a screenshot of its table. If the only source is a scan, use the PDF or image route and plan to review OCR output. A public webpage can be submitted by URL, while saved HTML belongs in the HTML converter. Choosing the right source avoids losing useful structure before conversion even starts. It also makes a later comparison easier because the Markdown can be checked against the document that actually contains the information. ## Reviewing the output Read the converted Markdown as a document, then compare the parts where its source format is most complicated. For a report, follow the reading order across columns and inspect table headers. For a presentation, confirm that notes belong to the slide they explain. For a workbook, check worksheet names and the position of numeric values. A result that looks tidy is not automatically complete. Keep the original open while you inspect the rendered preview and raw Markdown, and confirm that the end of the document is present. When you copy the result into another tool, review it there too: the destination may render tables, links or code differently from the preview on this page. ## Markdown for AI tools Convert to Markdown when you want a text representation that ChatGPT, Claude or your own retrieval pipeline can consume and that you can inspect yourself. Clear headings help preserve context, and supported tables keep relationships that would be difficult to recover from a loose list of numbers. The output still needs source review before you rely on it. Conversion does not verify a claim, choose which paragraphs are relevant or prevent a downstream model from making mistakes. Keep a reference to the source and provide the context needed for the question you are asking. For automated retrieval, evaluate chunking and search separately after you have verified the conversion itself. ## Anonymous trial limits The anonymous browser trial accepts a file up to 10 MB and 25 pages, with 3 conversions per client IP per hour and 10 a day. It is intended for checking a result before choosing a subscription. A new account does not create a monthly allowance of conversion credits. Lite, Pro and Max are the subscription plans for account conversion, saved history, batches and API use. Limits differ by plan, so check the pricing page before submitting a larger collection. A webpage must also be publicly accessible: URL conversion keeps the main content and renders JavaScript pages, but it does not log into a website. ## After conversion Keep the Markdown as an editable source for a note, a repository document or an ingestion process. If the next destination needs a different representation, the browser tools can render Markdown as HTML, remove formatting for plain text or open a print workflow for saving a PDF. Those steps change the representation of the output; they do not reconstruct the original document’s layout. For repeated input documents, move to Batch Convert or the developer API after checking representative samples. Keep the source name and a meaningful reference with the exported content so someone reviewing it later can identify where it came from. Conversion is most useful when the resulting text remains traceable and easy to check. ## FAQ ### What makes the output AI-ready? Conversion preserves headings, lists, tables, and reading order where possible, so the Markdown is predictable enough to chunk, embed, and retrieve without manual cleanup. ### Which file types can I convert? PDF, Word, PowerPoint, Excel, images, HTML, plain text, and public web pages by URL — the full list of file extensions is under Supported formats. The free public converter accepts files up to 10 MB. ### Do you support scanned documents? Image-based files and scanned documents are processed with OCR today. Recognition quality depends on the source image. ### How is Convert different from Batch Convert and the API? Convert is the single-document experience in the browser. Batch Convert runs many files at once, and the API runs the same conversion from your own code. They all share one conversion engine. ### Is this a document editor or a PDF layout tool? Neither. markitdown.ai converts documents into clean Markdown for AI and text workflows; it is not a visual PDF or layout editor. ## Sources - [CommonMark Spec](https://spec.commonmark.org/) - [RFC 7763: The text/markdown Media Type](https://www.rfc-editor.org/rfc/rfc7763) ## Related - [PDF to Markdown Converter](https://markitdown.ai/pdf-to-markdown) - [Batch convert to Markdown](https://markitdown.ai/batch-convert) - [Document Conversion APIs](https://markitdown.ai/developers) --- --- title: "Batch convert to Markdown" description: "Batch convert to Markdown is a bulk mode for paid plans that takes many files, folders, or ZIP archives, converts them in the background, and collects every result in one place." url: https://markitdown.ai/batch-convert updated: 2026-09-26 source: markitdown.ai --- # Batch convert to Markdown > Batch convert to Markdown is a bulk mode for paid plans that takes many files, folders, or ZIP archives, converts them in the background, and collects every result in one place. The anonymous trial accepts files up to 10 MB and 25 pages, with 3 conversions per IP per hour and 10 a day. Lite starts at $20 per month with files up to 100 MB. Pro is $50 per month; Max is $100 per month. - Add many files, a whole folder, or a ZIP archive at once - Per-file status: queued, running, done, failed - Track credits used and download results from one place ## Why convert in bulk Convert handles one file at a time. Batch Convert is for the point where converting documents becomes recurring work. - No manual loops — Add a stack of files, a folder, or a ZIP archive once instead of uploading and downloading them one at a time. - Runs in the background — Conversions process server-side, so you can close the tab and come back to the results. - Results in one place — Every output lands in your library, paired with its source file. - Predictable cost — See the credits a batch uses, so heavier files map to clearer usage. ## How a batch runs 1. Add your files — Select many documents, a folder, or a ZIP archive; the credit estimate appears before you start. 2. The batch runs — Files convert in the background through the same conversion engine as Convert. 3. Track progress — Watch each file move from queued to running to done — or surface the ones that failed. 4. Download or save — Collect the Markdown outputs, or keep them in your library for later reuse. ## Tracking built in A batch is only useful if you can see what happened. Every run keeps a record you can return to. - Batch history — Revisit past batches instead of re-running the same files. - Per-file status — Every file reports queued, running, done, or failed independently. - Retry failures — A single bad file does not sink the batch — handle it on its own. - Credits and usage — See how much each batch consumed and when limits reset. ## When to reach for Batch Convert Batch Convert is part of the paid plans. See the [pricing page](/pricing) for current limits. Need it in your own code instead of the UI? Send several files in one call to the [developer API](/developers). - RAG ingestion backfills — Convert an existing corpus before chunking and embedding. - Document migrations — Move large folders of PDFs and Office files into Markdown. - Recurring reports — Process a regular drop of documents on the same cadence. - Multi-file projects — Turn a whole project's worth of source documents into clean text at once. ## Preparing files Before you batch convert to Markdown, organize the collection so each output can be matched to its source. Use meaningful filenames and separate different revisions of the same document. Open a representative file from each input family and inspect a sample conversion before adding the whole collection. A set of scanned PDFs has different review needs from a set of editable Word documents, even when both become Markdown. Check that encrypted documents are unlocked and that the source files are readable. The batch workflow reduces repeated submission work; it does not remove the need to know what you are submitting or to recognize which version of a document produced a result. ## Completed batches A completed batch tells you that processing has reached an outcome, not that every source was converted successfully or that its text is ready for publication. Review each file’s status and keep failed items separate from successful outputs. A collection may finish with a mixture of useful results and files that need attention. Inspect representative outputs as well as the failure list, especially when your source material varies in quality. Record which documents you have checked in your own workflow before using the batch as a retrieval corpus or a published archive. Treat processing status, output quality and editorial approval as separate facts so a successful request is not mistaken for a reviewed document. ## Failed files Investigate the failed file before retrying it. A damaged source, an unsupported input or a password-protected document needs a source change; repeating the same submission may produce the same outcome. A temporary service or rate-limit error calls for a different response. Keep the conversion identifier and the visible error details so you can match the retry to its original attempt. Retry only the items that need another attempt and inspect their new status afterwards. Do not rerun an entire collection merely because one file failed. This keeps the review focused and makes it easier to explain which outputs belong to the original batch and which were produced after a correction. ## Batch limits A batch must fit both the subscription’s allowances and the submission method. A multipart API request can contain up to 100 files, but the whole request is limited to 100 MB. URL submissions accept up to 20 URLs per request. Those transport limits do not replace the per-file, page, storage and credit limits of Lite, Pro or Max. Files above a one-shot request limit need the documented upload workflow even when the subscription permits the file size. Plan concurrency controls how much work runs at once; it is not a promise that every file in a collection starts immediately. ## Building a knowledge base Keep each Markdown output connected to the source document and to the context in which it was collected. A file named report.md is difficult to distinguish from another report after both enter a search index. Your downstream workflow can retain a document title, source reference and revision information alongside the text. Review table-heavy, scanned and unusually long documents before splitting them into retrieval passages. If the batch contains repeated headers or web navigation, decide how to clean those elements in your ingestion process rather than assuming conversion removed them. A useful collection is one whose passages can be traced and verified, not simply a folder containing many Markdown files. ## Moving to the API Use Batch Convert when a person is organizing a collection and wants to inspect its progress and results in the web app. Use the document conversion API when a recurring process in your own system should submit inputs and receive outcomes without manual interaction. The API integration still needs a record for every source, a way to handle asynchronous completion and a policy for retries. Test that workflow with a small representative collection before running a larger one. Web and API use draw from the same account balance, so monitor combined usage. Automating submission should make repeated work easier to observe and recover, rather than hiding failures behind an apparently successful batch request. ## FAQ ### Which plans include Batch Convert? Batch Convert is available on paid plans. The free public converter handles a single file at a time. See the pricing page for current plan limits. ### What happens if one file fails? Each file is tracked on its own. A failed file is flagged so you can retry it without re-running the files that already succeeded. ### Which formats can I batch? The same formats as Convert — PDF, Word, PowerPoint, Excel, images, HTML, and text — added as files, whole folders, or ZIP archives. Batch Convert uses the same conversion engine. ### Where do the results go? Outputs land in your library paired with their source files, so you can preview, download, or reuse them after the batch finishes. ### Do I need to keep the tab open? No. Batch Convert runs server-side, so you can leave and return to the results from your batch history. ## Sources - [CommonMark Spec](https://spec.commonmark.org/) - [RFC 9110: HTTP Semantics](https://www.rfc-editor.org/rfc/rfc9110) ## Related - [Convert to Markdown](https://markitdown.ai/convert) - [markitdown.ai pricing for document conversion](https://markitdown.ai/pricing) - [Document Conversion APIs](https://markitdown.ai/developers) --- --- title: "Document Conversion APIs" description: "The document conversion API is an HTTPS API that takes a file, URL, or HTML in one POST /v1/convert/{name} call and returns Markdown." url: https://markitdown.ai/developers updated: 2026-09-26 source: markitdown.ai --- # Document Conversion APIs > The document conversion API is an HTTPS API that takes a file, URL, or HTML in one POST /v1/convert/{name} call and returns Markdown. The anonymous trial accepts files up to 10 MB and 25 pages, with 3 conversions per IP per hour and 10 a day. Lite starts at $20 per month with files up to 100 MB. Pro is $50 per month; Max is $100 per month. - One POST /v1/convert/{name} call per file, URL, or inline text - Markdown in the response within 60 seconds, or a conversion to poll - Same conversion engine as the web app ## How the document to Markdown API works Three steps from key to Markdown. Each call waits up to 60 seconds and answers with the result; longer or multi-file requests return 202 and a conversion to poll instead of holding the connection. **Request** ```bash # Upload a local file; Markdown comes back in the same response curl -X POST https://api.markitdown.ai/v1/convert/pdf \ -H "x-api-key: mdai_…" \ -F file=@nist-cswp-29-csf-2.0.pdf # Or convert a public web page curl -X POST https://api.markitdown.ai/v1/convert/url \ -H "x-api-key: mdai_…" \ -H "content-type: application/json" \ -d '{"url":"https://markitdown.ai/blog/why-pdfs-break-llms"}' # Took longer than the wait window? Long-poll the conversion curl "https://api.markitdown.ai/v1/conversions/?wait=100" \ -H "x-api-key: mdai_…" ``` 1. Create an API key — Create a key in Settings → Developers (needs an active subscription) and send it in the x-api-key header. 2. POST /v1/convert/{name} — Send a file, a public URL, or inline text to the endpoint for that format. One input kind per call. 3. Use the Markdown — Read the output from the response, or wait on /v1/conversions/{id} after a 202. ## Response and webhooks The response below illustrates the API contract using verified DOCX parser output; the conversion ID is a placeholder. A finished conversion returns the Markdown with the credits it cost. A 202 carries a Location header to the conversion; poll it with ?wait=, stream its events, or let a webhook tell you when it settles. **The response — or a conversion.completed webhook** ```text { "id": "cnv_3f7c…", "object": "conversion", "name": "word", "status": "completed", "files": [ { "outputs": [ { "format": "markdown", "content": "## Review Findings\n\nRisk ratings follow the standard High / Medium / Low scale defined in the vendor risk management policy.\n\n| Vendor | Risk Rating | Finding | Owner |\n| --- | --- | --- | --- |\n| CloudScan OCR | Medium | No documented data retention limit on uploaded scans | Security |\n| PayBridge | Low | SOC 2 Type II renewed; no open items | Finance |\n| LingoTrans API | High | Sub-processor list not disclosed on request | Legal |\n| ArchiveNow Storage | Medium | Encryption at rest confirmed; key rotation overdue | Security |\n" } ] } ], "usage": { "credits_charged": 1 } } POST https://your-app.example.com/webhooks { "type": "conversion.completed", "data": { "id": "cnv_3f7c…", "object": "conversion", "status": "completed" } } ``` Subscribe to `conversion.completed` and `conversion.failed` so you can react the moment a long or multi-file conversion settles instead of polling. Deliveries are signed with Standard Webhooks headers. ## Document parsing API and markdown conversion API Teams search for a document parsing API when they need structure out of files, and a markdown conversion API when the output has to land in a prompt or a vault. Both phrases describe this service. Format-specific landings such as the [pdf to markdown api](/pdf-to-markdown) and the [url to markdown api](/url-to-markdown) are anchors to those converters — they use the same `/v1/convert/{name}` endpoints. - Convert endpoints — POST /v1/convert/{name} — one endpoint per format — takes a file, URL, or text and returns Markdown, text, HTML, or per-page output. - API keys — Per-account keys sent via the x-api-key header, so conversion fits any server or script. - Multiple files and webhooks — Send many files to one convert call and get one conversion back; webhooks report each conversion.completed or conversion.failed. - Markdown Render — POST /v1/convert/markdown/html or /markdown/text turns Markdown back into HTML or plain text at no credit cost. ## Where teams wire it in - RAG ingestion — Convert documents inside ingestion jobs before chunking and embedding. - AI agents — Let agents turn arbitrary files into clean text mid-pipeline. - Content pipelines — Feed Markdown into a CMS, docs site, or knowledge base automatically. - Internal tools — Add document conversion to the products and back-office tools you already run. ## Auth, usage, and limits - API keys belong to your account, work on /v1 only, and are sent in the x-api-key header — the account needs an active Lite, Pro, or Max subscription - Usage uses the same credits as the web app — standard and OCR pages are 1 credit per page - Plan rate limits are Lite 20, Pro 40, and Max 60 requests per minute - Idempotency-Key on every POST /v1/convert/{name} call, so a retry never charges twice - Every v1 endpoint, request field, and response is documented in the generated [API reference](/docs/api); failures use problem+json with a `type` URI that resolves to the matching [error code page](/docs/errors) ## Handling completion Handle both inline completion and asynchronous processing in a document conversion API client. A successful submission can return a completed result, while HTTP 202 indicates that work continues after the response. Retain the conversion identifier and use the documented Location or conversion status endpoint before reading outputs. Do not index directly into an assumed first output simply because the HTTP request succeeded. A completed conversion can also contain file-level failures, so inspect each file in a multi-source request. Store enough context to connect every returned result to the source your application submitted. This keeps a slow conversion or a partial failure from silently becoming an empty document in your downstream system. ## Request vs plan limits The document conversion API applies limits to both requests and accounts. A multipart request can include up to 100 files with a combined limit of 100 MB. A URL request accepts up to 20 sources. Your subscription separately controls credits, storage, per-file limits and concurrency. A Pro file may be allowed by its subscription while being too large for one multipart request; use the documented chunked upload flow for that case. Do not work around a rejected request by assuming the highest plan removes transport limits. Read the error and use the input method intended for the source size. ## Keeping source context Keep your own source identifier next to the conversion identifier so a result can be traced back to the job that submitted it. The API metadata field carries caller-supplied data; it is not a promise to extract webpage titles, authors or publication dates. If your application needs those facts, collect and verify them in an appropriate separate step. Check whether an output is returned inline or through a referenced artifact before reading its text, and retain the original document when your workflow needs an audit trail. A document conversion API prepares content for the next stage, while your application remains responsible for associating that content with its source and deciding how long to retain it. ## Testing before launch Test more than one successful small file before integrating the document conversion API into a production workflow. Include an input that requires asynchronous completion, an unsupported or damaged file, a multi-source request with mixed outcomes and a response that your client must treat as a limit error. Verify that your own retry logic does not submit successful files again after a partial failure. Keep API credentials on your server and avoid placing them in browser code or public examples. Review the actual Markdown for representative source types before evaluating retrieval or model behavior downstream. A client that handles transport errors correctly still needs content-level checks to prove that the resulting documents are useful. ## FAQ ### How do I authenticate to the document to Markdown API? Create an API key in your account and send it in the `x-api-key` header. Keys are scoped to your account. ### What is the real conversion flow? POST /v1/convert/{name} with a file, a public URL, or text. The Markdown comes back in the same response when the document finishes within the wait window; otherwise you get 202 and a conversion to wait on with GET /v1/conversions/{id}?wait=. ### Is conversion synchronous? By default, up to the wait window: POST /v1/convert/{name} waits for a result and returns it inline, or returns 202 with a Location when the conversion cannot finish in time. Send Prefer: respond-async to skip waiting. ### What events can I subscribe to? Outbound webhooks support conversion.completed and conversion.failed, so you can react as soon as a conversion settles instead of polling. ### Which plans include API access? API access is available on paid plans. See the pricing page for current plan limits and credits. ### How is API usage billed? API conversions draw from the same credit balance as the web app. Standard pages and OCR pages cost 1 credit per page. ### Where do I find a pdf to markdown api or url to markdown api? Those phrases are format landings, not extra hosts. Use the pdf to markdown api notes on /pdf-to-markdown and the url to markdown api notes on /url-to-markdown, then call the same upload and convert endpoints documented here. ### What are the API key rate limits? Requests are rate limited per account: Lite 20, Pro 40, and Max 60 requests per minute. ## Sources - [RFC 9110: HTTP Semantics](https://www.rfc-editor.org/rfc/rfc9110) - [RFC 7763: The text/markdown Media Type](https://www.rfc-editor.org/rfc/rfc7763) ## Related - [Introduction](https://markitdown.ai/docs) - [PDF to Markdown Converter](https://markitdown.ai/pdf-to-markdown) - [markitdown.ai pricing for document conversion](https://markitdown.ai/pricing) --- --- title: "Markdown conversion, end to end." description: "Features is a catalog of how markitdown.ai converts PDFs, Office files, images, and HTML into structured Markdown for RAG and agents." url: https://markitdown.ai/features updated: 2026-09-26 source: markitdown.ai --- # Markdown conversion, end to end. > Features is a catalog of how markitdown.ai converts PDFs, Office files, images, and HTML into structured Markdown for RAG and agents. The anonymous trial accepts files up to 10 MB and 25 pages, with 3 conversions per IP per hour and 10 a day. Lite starts at $20 per month with files up to 100 MB. Pro is $50 per month; Max is $100 per month. - Parse: text layers read directly, scans and images through OCR - Viewer: source and Markdown side by side - Batch: ZIPs and URL lists with per-file status and retry - Library and account: history, usage, invoices, API keys ## Accurate Markdown from every page. Each file is routed by what is actually on its pages: text layers are read directly and scanned pages or images go through OCR, both at 1 credit per page. Headings, lists, tables and reading order come out as Markdown. - OCR for scans, photos and screenshots - AI figure enhancement for charts and figures on paid plans (3 credits per figure) - Cache results is optional: turn it off and a file is parsed fresh, without reusing a cached result ## Preview Markdown side by side. Open any conversion in the viewer: the source document on one side, rendered Markdown or raw source on the other. Spot a broken table before it reaches your pipeline. - Source and Markdown side by side - Preview or raw Markdown source, one click apart - Copy, or download the .md file ## Batch convert files to Markdown. Queue a ZIP of files or a list of URLs. Track every file, retry the ones that fail, and keep a history of each run. - Live progress with per-file status - Retry a failed file without re-running the rest - Batch history with credits per file ## Your Markdown library, on record. Every conversion is saved to your Library for your plan's history window — 30 days on Lite, 60 on Pro, 90 on Max. Usage, invoices and API activity live in Settings. - Library of source files paired with their Markdown - Usage by route and day, plus invoices - Scoped API keys, webhooks and an activity log ## Clean Markdown, down to the details. - Real Markdown tables — Rows, columns and header rows come out as GitHub-flavored pipe tables — not flattened runs of numbers — for PDFs, Office files, spreadsheets and HTML. - Cache control — Reuse matching results for repeated files, or turn caching off per run. - Figure descriptions — On paid plans, charts and figures can be described in words your model can reason about. - Documents, sheets, slides and web — PDF, Word, PowerPoint, Excel, CSV, EPUB, OpenDocument, images, HTML and URLs. - Auditable usage — Credits by route and day, plus invoices and an activity log. ## Test with a real sample Use a sample that reflects the work you actually need to do. Include a difficult table, a section boundary and source references rather than testing only a short paragraph. Compare the same document in the preview, the Markdown source and your intended destination. This separates a parsing difference from a display difference and gives you a practical basis for deciding whether an automated workflow is ready for a larger collection. ## Features by source type Choose Markdown conversion features by the source you actually receive. An editable Word document, an image-only PDF and a public web page expose different kinds of structure. A successful conversion should give you a readable representation of that source, but the checks you perform afterwards will differ. For a Word document, compare the heading outline and review comments. For a scan, inspect uncertain characters and table cells. For a web page, look for navigation or footer text that should not enter your downstream document. Trying a representative difficult input tells you more than converting only a simple paragraph. Keep the source and output together while you evaluate whether the conversion is useful for your task. ## Checking accuracy Evaluate Markdown conversion accuracy with passages you can compare directly against the source. Check a heading, a paragraph near a column boundary, a table with units and the final section of the document. A clean preview is useful, but it can hide an omitted qualification or a value under the wrong header. Switch to Markdown source when you need to inspect the exact text that a prompt or ingestion script will receive. The service prepares an output artifact; it does not independently establish the truth of the original material. Retain source references when you reuse extracted content, especially if another person will need to verify a quotation or follow a retrieved answer back to the document. ## Retrieval workflows Markdown conversion belongs before the application-specific work of splitting, indexing and retrieval. Headings and tables can give a splitter better boundaries than a flat text stream, but they do not choose an embedding model or guarantee that a search returns the right passage. Keep the document title and section context with the chunks your application creates. Test representative questions against answers you have verified in the source. If a table spans several chunks, check whether each retrieved piece still contains its headers and units. Tools such as LangChain and LlamaIndex can participate in your own ingestion workflow; the Markdown output is a reusable input to that workflow rather than a complete search system. ## Browser, batch and API Use the browser to inspect an individual result, Batch Convert to organize repeated files and the API to connect conversion to your own application. These entry points share the account’s conversion service and credit balance, but they solve different interaction needs. Browser review helps you see an output before automating it. A batch provides per-file outcomes that you can revisit. An API integration must handle authentication, request limits and asynchronous completion. Start automation only after reviewing the source types you expect to receive, and retain a way to investigate individual failures. Moving from one file to a large collection should change the way you track results, not remove the quality checks that made the first result useful. ## Available today Document, webpage and image parsing produce Markdown today. The browser also offers Markdown to HTML, plain text and a printable PDF. Browser printing is the available PDF output path; it is not a server-generated PDF download. Structured JSON extraction, audio and video transcription, and server-side PDF or DOCX output are planned capabilities. Do not build a current integration around a planned route or assume that OCR provides a complete description of a photograph or diagram. Read the specific converter page for the input and output you need, then verify a representative result. The product focuses on reusable content, so exact page-layout reconstruction and collaborative document editing are outside this workflow. ## FAQ ### What makes the output AI-ready? Conversion preserves headings, lists, tables, and reading order where possible, so the Markdown is predictable enough to chunk, embed, and retrieve without manual cleanup. ### Which file types can I convert? Common document formats include PDF, Word, PowerPoint, Excel, images and web content. The anonymous trial accepts files up to 10 MB and 25 pages, with 3 conversions per IP per hour and 10 a day. ### Can I use it for RAG or agent workflows? Yes. Clean Markdown is meant to be the input layer before chunking and embedding, and the API lets you run conversion inside ingestion jobs and agent pipelines. ### What is the difference between the web app and the API? The web app is for uploading, previewing, saving, and batching documents by hand. The API is for running the same conversion automatically inside your own product or tools. ### Can I batch convert documents? Yes. Batch conversion is available on paid plans so you can process repeated files without uploading and downloading one at a time. ### Do you support scanned documents? Image-based files and scanned documents are processed with OCR today. Recognition quality depends on the source image. ### Will schema JSON extraction be supported? It is a planned next layer, not a current feature. Markdown is the base artifact we build on first. ### Will Markdown translation be supported? Structure-preserving translation is on the roadmap as a planned next layer. ### Is this a document editor or a PDF layout tool? Neither. markitdown.ai converts documents into clean Markdown for AI and text workflows; it is not a visual PDF or layout editor. ## Sources - [CommonMark Spec](https://spec.commonmark.org/) - [GitHub Flavored Markdown Spec](https://github.github.com/gfm/) ## Related - [Convert to Markdown](https://markitdown.ai/convert) - [Batch convert to Markdown](https://markitdown.ai/batch-convert) - [Document Conversion APIs](https://markitdown.ai/developers) - [markitdown.ai pricing for document conversion](https://markitdown.ai/pricing) --- --- title: "markitdown.ai pricing for document conversion" description: "markitdown.ai pricing is a credit-metered plan table for Lite, Pro, and Max." url: https://markitdown.ai/pricing updated: 2026-09-26 source: markitdown.ai --- # markitdown.ai pricing for document conversion > markitdown.ai pricing is a credit-metered plan table for Lite, Pro, and Max. ## Plans and monthly credits markitdown.ai pricing is a credit-metered plan table for Lite, Pro, and Max. The anonymous trial accepts files up to 10 MB and 25 pages, with 3 conversions per IP per hour and 10 a day. Lite starts at $20 per month with files up to 100 MB. Pro is $50 per month; Max is $100 per month. Standard pages and OCR pages cost 1 credit per page; paid-plan accounts can use AI image understanding at 5 credits per image. Paid plans (Lite, Pro, Max) raise monthly credits, file size, pages per file, files per month, batch, and API access. Annual billing is 20% off and grants the year's credits up front. You can change plans or cancel anytime; upgrades are prorated. ## Estimating cost Include both routine conversions and occasional large documents in a pricing estimate. Keep an example of each input type you expect to process and check its observed usage after a successful run. A monthly total alone can hide a single file that exceeds a page or file-size allowance. Review your expected file count and storage needs alongside credits, then compare the plan that covers all of those requirements. ## Comparing plans Compare markitdown.ai pricing against the documents you actually process. Count pages, files and the size of your largest input separately; they describe different parts of usage. A collection of short documents can reach a file allowance before a page-heavy report reaches its credit allowance. Include web and API activity in the same estimate because they draw from one account balance. Check storage and history needs as well as conversion volume if you intend to return to source files later. The plan cards provide the current allowances, while the calculator is a planning aid. An estimate does not guarantee that a particular request fits every source, transport or account limit. ## What a credit measures Standard document parsing uses 1 credit per page, and OCR uses 1 credit per page. Credits describe metered processing rather than a promise that every file has the same cost. A long report and a short note can count as one file each while using different amounts of processing. Optional processing choices can add their own charges, so review the options and estimate before submitting work. The final account usage record is the place to check what a completed operation consumed. Keep file count, page count, storage and credits separate when you plan recurring work; combining them into a single informal limit makes it harder to understand which allowance you are approaching. ## Trial vs subscription The anonymous browser trial lets you inspect a conversion before subscribing. Its limits are 10 MB and 25 pages per file, with 3 conversions per client IP per hour and 10 a day. It does not create a recurring account credit allowance. Lite, Pro and Max are paid subscriptions for account conversions, batch processing, API access and saved history. Creating an account is not the same as activating a subscription. Use the public converter to evaluate a representative file, then compare the plan allowances against your expected workload. If a trial limit prevents you from testing a source, do not assume that repeatedly submitting the same file changes that limit. ## File limits Lite permits files up to 100 MB and 100 pages; Pro permits up to 200 MB and 200 pages. Max has no per-file size or page cap in its plan definition, while credits and storage still apply. A separate request-size limit also exists: one multipart request is limited to 100 MB in total. A larger file permitted by your plan needs the documented upload workflow. These are different constraints, so a plan’s larger file allowance should not be read as permission to send an arbitrarily large single HTTP request. Check the source, the plan and the input method together. ## Billing and history Monthly subscriptions provide the plan’s monthly credit allowance for the billing period. Annual subscriptions charge for the year and grant the corresponding annual credits up front. Keep the billing period distinct from history retention: Lite retains history for 30 days, Pro for 60 days and Max for 90 days. If you need a permanent copy of a reviewed result, download and retain it in your own workflow rather than treating saved history as an indefinite archive. Check your account’s displayed renewal and cancellation information when managing an existing subscription. The pricing page explains plan allowances; the account record shows the state of your particular subscription. ## FAQ ### What is a credit? Conversions are metered in credits. Standard pages and OCR pages cost 1 credit per page; paid-plan accounts can use AI image understanding at 5 credits per image. ### Do credits roll over? No. Monthly credits reset at the start of each billing period. ### How does annual billing work? Annual plans are billed once at 20% off and grant the full year's credits up front for the period. ### Can I change plans anytime? Yes. Upgrades apply immediately with prorated billing and top up the credit difference; downgrades adjust at the same time. You can cancel anytime. ### What happens when credits run out? Conversions pause until your next reset or an upgrade. ### Does the API use the same credits? Yes. API and web conversions draw from one credit balance. API access is included on paid plans. ### What are the file limits? Each plan sets a maximum file size, pages per file, and files per month. See the plan cards above. ### Can I try conversion before subscribing? The anonymous browser trial accepts files up to 10 MB and 25 pages, with 3 conversions per IP per hour and 10 a day. Subscriptions are Lite, Pro and Max; account creation does not grant monthly conversion credits. ### What happens when I delete a conversion or my account? Deleting a conversion removes it from your library at once and erases its stored files within an hour; a source file shared with another conversion is kept until that one is deleted too. Deleting your account erases all your stored source and result files. ## Sources - [RFC 7763: The text/markdown Media Type](https://www.rfc-editor.org/rfc/rfc7763) - [CommonMark Spec](https://spec.commonmark.org/) ## Related - [Markdown conversion, end to end.](https://markitdown.ai/features) - [Batch convert to Markdown](https://markitdown.ai/batch-convert) - [Document Conversion APIs](https://markitdown.ai/developers) --- --- title: "PDF to Markdown Converter" description: "PDF to Markdown is a converter that rebuilds multi-column reading order into structured Markdown for RAG. The public trial accepts files up to 10 MB." url: https://markitdown.ai/pdf-to-markdown updated: 2026-09-26 source: markitdown.ai --- # PDF to Markdown Converter > PDF to Markdown is a converter that rebuilds multi-column reading order into structured Markdown for RAG. The public trial accepts files up to 10 MB. ## How to convert PDF to Markdown - Upload a supported source in the PDF to Markdown converter. Choose a file you can compare with the result, and check the anonymous trial limits before starting. - Review PDF to Markdown output against the source. Check headings, table cells and the final passage before copying any content into another tool. - Copy the reviewed PDF to Markdown output for your next task. For recurring account conversions, saved history, batch or API access, choose a Lite, Pro or Max subscription. ## Why use PDF to Markdown - Reports as readable text — Turn a report’s paragraphs and supported structure into Markdown you can inspect, edit and keep beside the source PDF. - OCR for scanned pages — Image-only pages can be recognized with OCR. Compare small text, punctuation and numeric tables with the original scan. - Heading context — Retain recognizable sections for notes and retrieval workflows, while checking whether visually styled headings were interpreted correctly. - Tables to review — Supported tables become text tables. Check complex or merged cells before treating the exported values as evidence. - Preview before reuse — Compare the Markdown source with its rendered view before copying a passage into ChatGPT, Claude or a documentation repository. - A repeatable API route — Use the PDF conversion route in an ingestion script after reviewing representative outputs in the browser. ## Why PDF columns break language models - Columns interleave — A two-column report read top-to-bottom mixes left and right streams, so the model sees a sentence that never existed on the page. - Tables lose their grid — Copy-paste from a PDF turns a results table into a run of numbers with no header row for a downstream parser. - Scans have no text layer — Image-only pages carry pixels, not glyphs, so scanned PDFs are processed with OCR today. ## What this PDF converter keeps - Reading order reconstruction — PDF conversion attempts to reconstruct a logical reading order. Review column transitions against the original, especially for complex reports and papers. - Tables and headings — Heading levels and tabular data are preserved where the PDF exposes structural metadata; flat-text PDFs produce flat-text output. - Scanned and image-only PDFs — PDFs without a text layer (scanned pages and image inserts) are processed with OCR today; accuracy depends on scan quality. - Password-protected files — Encrypted PDFs cannot be parsed; remove the password before uploading. ## PDF to Markdown in real workflows - Research notes — Extract a paper into editable notes, verify quoted passages and retain the PDF reference for later review. - Report ingestion — Prepare reports for chunking while keeping table headers, section context and a link to the original document. - Policy libraries — Convert a reviewed policy into searchable text and preserve the published PDF as the authoritative version. - Document evaluation — Compare representative PDFs before automating a collection, including scans, columns and a difficult table. ## PDF to Markdown API Use POST /v1/convert/pdf with an x-api-key header and a supported source file. A completed request can return outputs inline; HTTP 202 means conversion is still running. Read the Location response header and poll it before accessing file outputs. A multipart request accepts at most 100 files; subscription and total request limits also apply. Keep API keys in your server environment. API access requires a Lite, Pro or Max subscription. ```bash curl -i "https://api.markitdown.ai/v1/convert/pdf" \ -H "x-api-key: $MARKITDOWN_API_KEY" \ -F "file=@source.pdf" # If HTTP 202, GET the Location URL with the same API key. ``` ## Reviewing the output Check PDF to Markdown output against the original PDF before you use it as evidence. Start with a page containing columns, a footnote and a table rather than an easy title page. Read the last sentence of one column and the first sentence of the next: they should form the same sequence a person follows in the PDF. Then compare a table header, a negative value and a unit. A readable Markdown preview can still contain an OCR substitution or a misplaced cell. If the document informs a financial or technical decision, keep the PDF alongside the Markdown and verify the passage you intend to quote. Conversion makes the text easier to work with; it does not certify the source or its transcription. ## What changes PDF to Markdown conversion replaces visual positioning with a text structure. A large heading may become a heading marker, paragraphs become paragraphs, and a supported table becomes rows separated by pipes. Page geometry, decorative rules, font choices and the original line wrapping are not the output contract. This is useful when you want to edit an extract or build a retrieval index, but it is unsuitable when the goal is to reproduce a brochure exactly. Keep the original PDF for its authoritative layout. Where a layout cannot be represented by a simple Markdown table, check whether a short explanation or a manually reviewed table is the clearer downstream representation. ## RAG sources Use the PDF title and meaningful section headings to retain context when splitting Markdown into passages. A chunk headed only Results may be ambiguous after it leaves the document; your ingestion code can retain the document name and the parent heading with that passage. Keep a table header together with its rows and avoid splitting a list halfway through a qualification. Store the original PDF reference separately so a retrieved answer can lead a reader back to the source. LangChain and LlamaIndex can consume text from your own ingestion process, but choosing a splitter, embedding model and retrieval policy remains your application’s job. Test retrieval with questions whose answers you have checked in the PDF. ## Choosing the PDF Prefer the original PDF export when it contains readable text. A screenshot of that PDF introduces another image recognition step and can lose small characters before conversion begins. For scanned PDFs, straighten rotated pages and use a version in which punctuation and table borders remain legible at normal zoom. Remove a password before uploading an encrypted PDF. If only part of a long report is relevant, prepare the relevant source pages and keep their original reference in your notes. Repeating the same damaged scan is unlikely to fix missing detail; improve the source first, then compare the new PDF to Markdown output with the previous result. ## Chunk boundaries Retrieval quality depends on chunk boundaries that match the document a human would read. This converter reorders columns and keeps heading levels so a splitter can emit passages instead of interleaved fragments from a two-column paper. - Use the Markdown as the artifact you chunk and embed, not a flattened text dump. - Standard pages and OCR pages are billed at 1 credit per page on signed-in plans. - Password-protected files must be unlocked before upload; encrypted PDFs are rejected. ## Example: NIST Cybersecurity Framework 2.0, page 15 Source: https://nvlpubs.nist.gov/nistpubs/CSWP/NIST.CSWP.29.pdf ```markdown ## Appendix A. CSF Core This appendix describes the Functions, Categories, and Subcategories of the CSF Core. Table 1 lists the CSF 2.0 Core Function and Category names and unique alphabetic identifiers. Each Function name in the table is linked to its portion of the appendix. The order of Functions, Categories, and Subcategories of the Core is not alphabetical; it is intended to resonate most with those charged with operationalizing risk management within an organization. The numbering of the Subcategories is intentionally not sequential; gaps in numbering indicate CSF 1.1 Subcategories that were relocated in CSF 2.0. **Table 1. CSF 2.0 Core Function and Category names and identifiers** | Function | Category | Category Identifier | | --- | --- | --- | | **Govern (GV)** | Organizational Context | GV.OC | | | Risk Management Strategy | GV.RM | | | Roles, Responsibilities, and Authorities | GV.RR | | | Policy | GV.PO | | | Oversight | GV.OV | | | Cybersecurity Supply Chain Risk Management | GV.SC | | **Identify (ID)** | Asset Management | ID.AM | | | Risk Assessment | ID.RA | | | Improvement | ID.IM | | **Protect (PR)** | Identity Management, Authentication, and Access Control | PR.AA | | | Awareness and Training | PR.AT | | | Data Security | PR.DS | | | Platform Security | PR.PS | | | Technology Infrastructure Resilience | PR.IR | | **Detect (DE)** | Continuous Monitoring | DE.CM | | | Adverse Event Analysis | DE.AE | ``` ## FAQ ### How does PDF to Markdown work? PDF to Markdown converts the supported source into readable Markdown for review and reuse. The anonymous trial accepts up to 10 MB and 25 pages per file, with 3 conversions per IP per hour and 10 a day. Compare the result with the source before using it. ### What is the maximum PDF size? The public converter accepts PDF files up to 10 MB. Lite supports files up to 100 MB; Pro and Max have higher limits. ### Does it handle scanned PDFs? Yes. Image-based PDFs are processed with OCR today; clean, high-resolution scans produce the best results. ### Are multi-column research papers handled? Yes. PDF to Markdown attempts to reconstruct reading order. Compare column transitions with the source, because complex layouts can still need correction. ### Will financial tables survive? Tables in PDFs that carry structural metadata are converted to Markdown tables. Flattened-image tables require OCR to recover. ### Can I convert password-protected PDFs? No. Encrypted PDFs cannot be parsed. Remove the password before uploading. ### Is the output good enough to chunk for RAG? The Markdown preserves headings, lists, and tables where possible so chunking and embedding stay predictable compared with copy-paste. ### Can I convert PDFs from code? Yes. Use an API key to POST the file straight to /v1/convert/pdf and read the Markdown back in the response. See /developers. ## Sources - [ISO 32000-2 (PDF 2.0)](https://www.iso.org/standard/75839.html) - [CommonMark Spec](https://spec.commonmark.org/) ## Related - [Markdown converter for documents and web pages.](https://markitdown.ai/) --- --- title: "Word to Markdown Converter" description: "Word to Markdown is a converter that maps DOC and DOCX heading styles to Markdown and appends review comments. Try files up to 10 MB without signing in." url: https://markitdown.ai/word-to-markdown updated: 2026-09-26 source: markitdown.ai --- # Word to Markdown Converter > Word to Markdown is a converter that maps DOC and DOCX heading styles to Markdown and appends review comments. Try files up to 10 MB without signing in. ## How to convert Word to Markdown - Upload a supported source in the Word to Markdown converter. Choose a file you can compare with the result, and check the anonymous trial limits before starting. - Review Word to Markdown output against the source. Check headings, table cells and the final passage before copying any content into another tool. - Copy the reviewed Word to Markdown output for your next task. For recurring account conversions, saved history, batch or API access, choose a Lite, Pro or Max subscription. ## Why use Word to Markdown - A portable document outline — Word headings become a text outline you can review without carrying font and page-layout decisions into your next document. - DOCX review context — Supported review comments are collected as text. Keep the Word original when a comment’s precise location matters. - Editable Markdown — Use the output in a Markdown editor or version-controlled repository, then review it in your destination renderer. - Lists and tables — Preserve supported lists and tabular content while checking nesting, headers and cells against the editable Word source. - One source for reuse — Start with the original Word file instead of printing it to PDF and reinterpreting the resulting page layout. - Browser and API — Review one document interactively, then use the Word API route for the same input family in an automated workflow. ## Why Word styles pollute prompts - Style noise — Tracked changes, fonts, and inline styling leak into copy-pasted text and pollute prompts. - Embedded tables — Tables and lists lose their structure when a proposal is pasted as plain text. - Comments get lost — Review comments are preserved in a dedicated `## Comments` section instead of being silently dropped. ## What this Word converter keeps - Headings and lists — Word heading styles (Heading 1–6) map directly to Markdown `#` levels; bulleted and numbered lists are preserved. - Tables — Tables in the document convert to Markdown pipe tables with the original column structure intact. - Tracked changes and comments — Inserted text is included and deleted text is omitted. Review comments are appended under `## Comments`; their inline anchors are not preserved. - Embedded images — Embedded images are referenced in the output but not inlined as base64; the Markdown will note their position. ## Word to Markdown in real workflows - Documentation migration — Move a reviewed Word draft into Markdown, checking headings, links and comments before publishing it. - Editorial handoff — Share text that another editor can work on while retaining the DOCX for revision and comment context. - Knowledge base inputs — Prepare internal documents for search with meaningful section names and references to source versions. - Repository workflows — Keep reviewed documentation as text so contributors can discuss a change in a focused diff. ## Word to Markdown API Use POST /v1/convert/word with an x-api-key header and a supported source file. A completed request can return outputs inline; HTTP 202 means conversion is still running. Read the Location response header and poll it before accessing file outputs. A multipart request accepts at most 100 files; subscription and total request limits also apply. Keep API keys in your server environment. API access requires a Lite, Pro or Max subscription. ```bash curl -i "https://api.markitdown.ai/v1/convert/word" \ -H "x-api-key: $MARKITDOWN_API_KEY" \ -F "file=@source.docx" # If HTTP 202, GET the Location URL with the same API key. ``` ## Styles to headings Word to Markdown works best when the document uses paragraph styles to express its outline. A paragraph styled as a heading carries a different signal from ordinary text that merely looks large and bold. After conversion, inspect the heading sequence in the Markdown source and compare it with the Word navigation outline. A section should not become a new document title just because it starts on a new page. If the source relies on manual formatting, review the result before publishing it into a documentation repository. This is also a useful moment to replace purely visual distinctions with clear section names that make sense in plain text. ## Reviewing the output Review DOCX to Markdown output in the destination where it will actually be read. A table may be valid Markdown but too wide for a small screen, and a link may be correct while its label gives the reader too little context. Open a representative document with headings, nested lists and reviewer comments before migrating an entire collection. Compare the content after the last table as well as the beginning of the file, because important conclusions often sit near the end. Preserve the source DOCX with the exported Markdown so editors can investigate a difference without guessing which version produced the output. A conversion is a starting artifact for publication, not an automatic editorial approval. ## Comments and revisions Word to Markdown exports the readable document and collects supported DOCX review comments as text. A comments section is useful for retaining feedback, but it does not recreate the Word review interface. Do not assume that an exported comment remains attached to the exact character range it annotated. Review decisions that depended on an inline anchor still need the original document. Likewise, Markdown is not a tracked changes format: readers should not expect to accept or reject revisions inside the exported file. Resolve disputed wording in the source before publishing a final version, then check that the exported paragraphs contain the intended accepted text. ## Word vs PDF Use Word to Markdown when the editable Word document is your source of truth and the next step is editing, version control or reuse in a text workflow. Converting the same document to PDF first can turn its useful paragraph structure into page layout that then has to be interpreted again. PDF conversion is appropriate when the PDF is the only authoritative version you have, such as a signed or published report. For a documentation migration, start with the Word source where available, keep images and references under review, and compare the finished Markdown in the target renderer. Choose the route by the source you possess, not merely by the filename you want at the end. ## Docs workflows Editorial documents carry tracked changes and review balloons that disappear in a plain-text paste. This converter exports the accepted view and collects comments under a dedicated heading so a writer can still see the review thread. - Heading 1–6 styles map to Markdown `#` levels. - Inserted text is kept; deleted text is omitted. - Inline comment anchors are not preserved — only the comment text under `## Comments`. ## Example: vendor security review, Q3 2026 Source: markitdown.ai example: vendor-security-review-q3-2026.docx ```markdown ## Review Findings Risk ratings follow the standard High / Medium / Low scale defined in the vendor risk management policy. | Vendor | Risk Rating | Finding | Owner | | --- | --- | --- | --- | | CloudScan OCR | Medium | No documented data retention limit on uploaded scans | Security | | PayBridge | Low | SOC 2 Type II renewed; no open items | Finance | | LingoTrans API | High | Sub-processor list not disclosed on request | Legal | | ArchiveNow Storage | Medium | Encryption at rest confirmed; key rotation overdue | Security | ## Remediation Timeline 1. Send updated data processing addendum to LingoTrans API and request the sub-processor list 2. Confirm ArchiveNow Storage completes key rotation before the next billing cycle 3. Re-run the questionnaire for any vendor rated High within 30 days of remediation 4. Report closed items to the security review board at the end of Q4 LingoTrans API remains the highest-priority item on this list. Escalate to Legal if the sub-processor list is not provided within two weeks. ## Comments - **Priya Nair**: Confirm with Legal whether we can suspend new uploads to LingoTrans until this is resolved. ``` ## FAQ ### How does Word to Markdown work? Word to Markdown converts the supported source into readable Markdown for review and reuse. The anonymous trial accepts up to 10 MB and 25 pages per file, with 3 conversions per IP per hour and 10 a day. Compare the result with the source before using it. ### What is the maximum Word file size? The public converter accepts Word files up to 10 MB. The anonymous trial also allows up to 25 pages and 3 conversions per IP per hour and 10 a day. ### Will heading styles become Markdown headings? Yes. Heading hierarchy, lists, and tables are preserved where the document exposes those styles. ### Are tracked changes included in the output? The accepted view is exported: inserted text is included and deleted text is omitted. Review comments appear under `## Comments`. ### What happens to embedded images in a DOCX? Embedded images are referenced by position in the output but not inlined as base64 data. ### Does it handle password-protected Word files? No. Password-protected Word files cannot be parsed. Remove the document password before uploading. ### Do I need an account to preview a proposal? No. The anonymous trial lets you preview and copy a result. A Lite, Pro or Max subscription enables account conversions, saved history, batch and API access. ### Can I automate Word conversion? Yes. POST the .docx to /v1/convert/word with an API key and read the Markdown from the response. See /developers. ## Sources - [ECMA-376 Office Open XML](https://ecma-international.org/publications-and-standards/standards/ecma-376/) - [CommonMark Spec](https://spec.commonmark.org/) ## Related - [Markdown converter for documents and web pages.](https://markitdown.ai/) --- --- title: "PPT to Markdown Converter" description: "PPT to Markdown is a converter that turns each slide into a Markdown section and appends speaker notes. Standard pages cost 1 credit per page." url: https://markitdown.ai/ppt-to-markdown updated: 2026-09-26 source: markitdown.ai --- # PPT to Markdown Converter > PPT to Markdown is a converter that turns each slide into a Markdown section and appends speaker notes. Standard pages cost 1 credit per page. ## How to convert PPT to Markdown - Upload a supported source in the PPT to Markdown converter. Choose a file you can compare with the result, and check the anonymous trial limits before starting. - Review PPT to Markdown output against the source. Check headings, table cells and the final passage before copying any content into another tool. - Copy the reviewed PPT to Markdown output for your next task. For recurring account conversions, saved history, batch or API access, choose a Lite, Pro or Max subscription. ## Why use PPT to Markdown - Slide boundaries — Turn a presentation into a readable sequence and retain slide context when building written notes from the deck. - Speaker note context — Check extracted notes alongside the slide they explain, and review private rehearsal comments before sharing the export. - Text that can be searched — Use slide text in a search or retrieval workflow while keeping a reference back to the original presentation. - Supported chart data — Review supported chart and table content against the slide; visual relationships may still require a human explanation. - Markdown for handouts — Create a text starting point for training material, then add context for diagrams and presentation-only references. - API automation — Send PowerPoint files through the dedicated conversion route and record each result with the source deck name. ## Why slide text boxes scatter - Slides fragment — Text boxes scatter across each slide, producing disjointed fragments instead of a readable outline. - Chart values vanish — Numbers inside embedded charts are easy to lose; structure-aware parsing keeps the data when the deck stores it. - Notes are ignored — Speaker notes and grouped shapes need careful handling to end up as usable Markdown. ## What this PPT converter keeps - One section per slide — Each slide becomes its own Markdown section with the slide title as a heading, keeping the deck's structure intact. - Speaker notes — Speaker notes are extracted and appended after each slide's content, clearly separated from the main text. - Charts and data — Embedded charts are rendered using their underlying data values and labels where available; purely decorative chart images may not yield structured output. - Legacy .ppt files — Both legacy .ppt and modern .pptx are supported, though older binary .ppt files may have lower fidelity than .pptx. ## PPT to Markdown in real workflows - Training notes — Make a deck searchable and add written explanation where a slide depends on a live presenter. - Research briefings — Extract a briefing’s text while keeping slide references for figures and claims that need visual review. - Knowledge sharing — Turn supported slide content into a first draft for a team knowledge base, then review notes for suitability. - Presentation archives — Index a collection of decks with source names so readers can retrieve the original presentation when needed. ## PPT to Markdown API Use POST /v1/convert/powerpoint with an x-api-key header and a supported source file. A completed request can return outputs inline; HTTP 202 means conversion is still running. Read the Location response header and poll it before accessing file outputs. A multipart request accepts at most 100 files; subscription and total request limits also apply. Keep API keys in your server environment. API access requires a Lite, Pro or Max subscription. ```bash curl -i "https://api.markitdown.ai/v1/convert/powerpoint" \ -H "x-api-key: $MARKITDOWN_API_KEY" \ -F "file=@source.pptx" # If HTTP 202, GET the Location URL with the same API key. ``` ## Reviewing the output PPT to Markdown turns a presentation into a linear reading document. Slide titles provide boundaries, while supported text and speaker notes supply the explanation inside those boundaries. Position, animation and the order in which a presenter reveals objects are not the same thing as reading order. A diagram whose meaning depends on arrows may need a human explanation even when all of its labels appear in the Markdown. Read the export without looking at the slides: if a sentence says as shown here but the relationship is missing, add context in your downstream notes. Keep the original PowerPoint deck for visual review and for anyone who needs to present the material again. ## Checking speaker notes Speaker notes can carry the actual explanation behind a brief slide title, so compare a slide with notes against its PPT to Markdown output. Confirm that the notes remain associated with the intended slide and that a repeated footer has not become part of the argument. Notes may contain rehearsal reminders or material that was never intended for a public audience. Review those passages before sharing the Markdown with customers or adding it to a knowledge base. A searchable export makes hidden context easier to find, which also makes an editorial check more important. The converter preserves supported content; deciding which notes belong in a published document remains your responsibility. ## Training material Use PPT to Markdown to make the text of a training deck available for search, study notes or a first draft of a written guide. Keep the slide sequence when it explains a process, but add descriptive section names in the downstream document when the original titles are vague. A slide called Example is understandable during a talk and much less useful as an isolated search result. Check that every extracted chart description matches the visible chart rather than treating a chart title as its full meaning. In a retrieval index, retain a reference to the deck and slide so someone can inspect the original visual when a text answer needs clarification. ## Choosing a deck Choose the original editable PowerPoint file when possible. A deck made entirely from screenshots contains far less editable structure than a deck with real text boxes and tables. If you have only a scanned handout, use the PDF or image converter appropriate to that source and review the OCR result. For an editable deck, test a slide with a table, a slide with a chart and a slide with speaker notes before processing a large library. Do not interpret successful conversion as proof that every visual relationship survived. The useful outcome is text that accurately represents the deck’s content, together with enough source context to resolve what text alone cannot show. ## Notes and chart values A deck is more than titles on a canvas. Speaker notes often hold the argument, and embedded charts hold the numbers. This converter appends notes after each slide and renders chart data when the file stores values and labels. - Each slide becomes its own Markdown section titled from the slide title. - Decorative chart images without underlying data may not yield a table. - Animations and transitions are presentation-only and are not represented. ## Example: Q3 2026 onboarding funnel review deck Source: markitdown.ai example: q3-2026-onboarding-funnel-review.pptx ```markdown ## What changed this quarter - Removed the mandatory phone verification step - Shortened the free-trial signup form from 6 fields to 3 - Added a progress indicator to the document upload step - Moved pricing plan selection after the first successful conversion > Each of these shipped independently between weeks 2 and 9; the phone verification removal alone accounts for roughly a third of the lift. Funnel conversion by step | Step | Q2 2026 | Q3 2026 | | --- | --- | --- | | Signup started | 100% | 100% | | Signup completed | 61% | 74% | | First file converted | 44% | 52% | | Upgraded to paid | 3.1% | 3.8% | > Signup-completed is the biggest jump — that's the form-shortening change. First-file-converted improved less than expected; that's next quarter's focus. ``` ## FAQ ### How does PPT to Markdown work? PPT to Markdown converts the supported source into readable Markdown for review and reuse. The anonymous trial accepts up to 10 MB and 25 pages per file, with 3 conversions per IP per hour and 10 a day. Compare the result with the source before using it. ### What is the maximum deck size? The public converter accepts PowerPoint files up to 10 MB. The anonymous trial also allows up to 25 pages and 3 conversions per IP per hour and 10 a day. ### Are charts and tables included from slides? Chart titles and tabular data are captured where possible and appended as Markdown. Decorative chart images without underlying data may not produce structured output. ### Are speaker notes extracted from a deck? Yes. Speaker notes are extracted and appended after each slide's text content in the Markdown output. ### How are slides structured in the Markdown? Each slide becomes a section with the slide title as a Markdown heading, followed by the slide body and any speaker notes. ### Does it handle animations or transitions? No. Animations, transitions, and other presentation-only effects are not represented in the Markdown output. ### Is sign-up required to preview a deck? No. The anonymous trial lets you preview and copy a result. Account conversions and saved history require a Lite, Pro or Max subscription. ### Can I convert decks from a script? Yes. POST the .pptx to /v1/convert/powerpoint with an API key; the response carries the Markdown. See /developers. ## Sources - [ECMA-376 Office Open XML](https://ecma-international.org/publications-and-standards/standards/ecma-376/) - [CommonMark Spec](https://spec.commonmark.org/) ## Related - [Markdown converter for documents and web pages.](https://markitdown.ai/) --- --- title: "Excel to Markdown Converter" description: "Excel to Markdown is a converter that emits one Markdown table per worksheet with the sheet name as a heading. The anonymous trial also allows up to 25 pages and 3 conversions per IP per hour and 10 a day." url: https://markitdown.ai/excel-to-markdown updated: 2026-09-26 source: markitdown.ai --- # Excel to Markdown Converter > Excel to Markdown is a converter that emits one Markdown table per worksheet with the sheet name as a heading. The anonymous trial also allows up to 25 pages and 3 conversions per IP per hour and 10 a day. ## How to convert Excel to Markdown - Upload a supported source in the Excel to Markdown converter. Choose a file you can compare with the result, and check the anonymous trial limits before starting. - Review Excel to Markdown output against the source. Check headings, table cells and the final passage before copying any content into another tool. - Copy the reviewed Excel to Markdown output for your next task. For recurring account conversions, saved history, batch or API access, choose a Lite, Pro or Max subscription. ## Why use Excel to Markdown - Named worksheets — Worksheet names help separate the exported tables so readers can identify the part of the workbook they are reading. - Readable table syntax — Review rows and headers as Markdown rather than pasting a spreadsheet into a prompt as an unlabeled block of values. - Text version control — Use a reviewed table in repository documentation where a text diff can show which entries have changed. - Original workbook context — Keep the workbook for live formulas and editing. The Markdown output is a text representation for reuse. - Fewer manual transcription steps — Convert the worksheet before editing your report, then check dates, numbers, empty cells and units against the source. - API for repeated exports — Use the Excel route to process saved workbook inputs in your own reporting workflow, with per-file result checks. ## Why spreadsheet paste loses the grid - Tables collapse — Spreadsheet data pasted as plain text loses its rows and columns and becomes useless downstream. - Multiple sheets — Workbooks with several sheets need each one labelled, not merged into one blob. - Merged cells — Merged and formatted cells distort naive extraction; the converter flattens them instead of inventing extra columns. ## What this Excel converter keeps - One section per worksheet — Each worksheet becomes its own Markdown section with the sheet name as a heading, keeping multi-tab workbooks readable. - Cells as Markdown tables — Cell data is rendered as a standard Markdown pipe table with the first row treated as the header row. - Merged cells — Merged cells are flattened: the value appears in the top-left cell of the merged range and the remaining cells are left empty. - Very wide sheets — Sheets with many columns produce wide Markdown tables that may wrap in narrow editors; consider trimming unused columns before uploading. ## Excel to Markdown in real workflows - Documentation tables — Publish a reviewed worksheet table in a Markdown document with units and source context attached. - Research appendices — Keep a readable table next to an analysis while preserving the workbook for calculations and provenance. - Reporting handoffs — Give a writer structured text to work with instead of asking them to retype spreadsheet cells. - Repository reference data — Review table changes as text while retaining the source workbook and a clear refresh procedure. ## Excel to Markdown API Use POST /v1/convert/excel with an x-api-key header and a supported source file. A completed request can return outputs inline; HTTP 202 means conversion is still running. Read the Location response header and poll it before accessing file outputs. A multipart request accepts at most 100 files; subscription and total request limits also apply. Keep API keys in your server environment. API access requires a Lite, Pro or Max subscription. ```bash curl -i "https://api.markitdown.ai/v1/convert/excel" \ -H "x-api-key: $MARKITDOWN_API_KEY" \ -F "file=@source.xlsx" # If HTTP 202, GET the Location URL with the same API key. ``` ## What a workbook keeps Excel to Markdown turns supported worksheet data into readable text tables, with sheet names helping readers identify the source. It does not turn Markdown into a spreadsheet calculation engine. A number in the output is a value to review, not a live cell that updates when another value changes. Keep the workbook when formulas, filters, validation rules or editing interactions matter to the next task. For a report or a repository document, Markdown tables are easier to read and compare as text. For ongoing financial modelling, the original spreadsheet remains the right working format. Decide whether your goal is to explain the data or continue calculating with it before choosing the export. ## Preparing the table Use descriptive worksheet names and a clear header row before converting Excel to Markdown. A worksheet called Sheet1 with repeated unlabeled totals gives a reader little context once it becomes a text table. If several unrelated tables share a worksheet, review where one table ends and the next begins in the export. Merged headings, decorative empty rows and color-only status indicators deserve particular attention because Markdown has a simpler structure. Preserve units in text rather than relying solely on cell formatting. A value marked red in Excel does not explain itself in a plain text document; include the meaning of that status in your downstream description when it matters. ## Reviewing the output Compare an Excel to Markdown table with representative cells in the workbook, including dates, percentages, negative numbers and blanks. Check a value from the first row and one near the end, then confirm that both still sit beneath the correct header. A blank cell and a numeric zero are different facts and should not be casually substituted in a downstream cleanup step. If the workbook depends on calculated results, save a current copy from your spreadsheet application before conversion and verify that the displayed result is what you intend to share. Markdown is convenient for reading numbers, but the presence of a well-formed table is not evidence that a calculation is correct. ## Tables in a repository Excel to Markdown is useful for publishing a reviewed table in project documentation, release notes or a research report. A text diff makes it possible to see which row changed without opening a spreadsheet application. Keep the source workbook and describe when the exported table was produced so contributors know how to refresh it. Wide worksheets can be awkward on phones, so consider publishing a focused table in your downstream document and linking to the complete source. The converter does not choose which columns are relevant to your reader. Review the exported table as a communication artifact, and retain the workbook for analysis that needs formulas or spreadsheet interactions. ## One table per sheet A workbook is a stack of grids, not one blob. Each worksheet becomes its own section so a model can cite Sheet1 separately from Sheet2. Merged cells flatten: the value stays in the top-left cell of the range. - The first row of each sheet is treated as the header row of the pipe table. - Computed cell values are captured; underlying formulas are not. - Embedded charts and images inside the spreadsheet are not converted. ## Example: Q3 2026 revenue by region workbook Source: markitdown.ai example: q3-2026-revenue-by-region.xlsx ```markdown ## Q3 Revenue by Region | Region | Q1 2026 | Q2 2026 | Q3 2026 | YoY Growth | | --- | --- | --- | --- | --- | | North America | 812400 | 845100 | 901200 | 11.0% | | EMEA | 431900 | 468200 | 512700 | 18.7% | | APAC | 265300 | 298800 | 341500 | 28.7% | | LATAM | 74200 | 81900 | 93100 | 25.5% | | Total | 1583800 | 1694000 | 1848500 | | ## EMEA Account Detail | Account | Plan | Seats | Monthly Revenue | Renewal Date | | --- | --- | --- | --- | --- | | Nordwind Logistics GmbH | Max | 40 | 4800 | 2026-11-01 | | Iberia Consulting SL | Pro | 12 | 960 | 2026-12-15 | | Baltic Ferries AB | Pro | 8 | 640 | 2027-01-20 | | Meridian Legal LLP | Lite | 5 | 100 | 2026-10-05 | ``` ## FAQ ### How does Excel to Markdown work? Excel to Markdown converts the supported source into readable Markdown for review and reuse. The anonymous trial accepts up to 10 MB and 25 pages per file, with 3 conversions per IP per hour and 10 a day. Compare the result with the source before using it. ### Does it support XLS and XLSX? Yes, both legacy .xls and modern .xlsx are supported, along with .ods. ### What is the maximum workbook size? The public converter accepts Excel files up to 10 MB. The anonymous trial also allows up to 25 pages and 3 conversions per IP per hour and 10 a day. ### How are merged cells handled in the table? Merged cells are flattened: the value is placed in the top-left cell of the merged range and the remaining cells are empty in the Markdown table. ### Are formulas preserved from the workbook? No. The converter captures the computed cell values, not the underlying formulas. ### What about charts and images inside the spreadsheet? Embedded charts and images are not converted to Markdown; only cell data is extracted. ### Do I need to sign in to preview a sheet? No. Preview and copy are available through the anonymous trial. Account conversions require a Lite, Pro or Max subscription. ### Can I convert spreadsheets from an ingestion job? Yes. POST the workbook to /v1/convert/excel from the job, or send several workbooks in one call. See /developers. ## Sources - [ECMA-376 Office Open XML](https://ecma-international.org/publications-and-standards/standards/ecma-376/) - [GitHub Flavored Markdown Spec](https://github.github.com/gfm/) ## Related - [Markdown converter for documents and web pages.](https://markitdown.ai/) --- --- title: "EPUB to Markdown Converter" description: "EPUB to Markdown is a converter that turns an ebook's chapters into structured Markdown, billed at 1 credit per 1,800 characters of output. The anonymous trial accepts EPUB files up to 10 MB and 25 pages." url: https://markitdown.ai/epub-to-markdown updated: 2026-09-28 source: markitdown.ai --- # EPUB to Markdown Converter > EPUB to Markdown is a converter that turns an ebook's chapters into structured Markdown, billed at 1 credit per 1,800 characters of output. The anonymous trial accepts EPUB files up to 10 MB and 25 pages. ## How to convert EPUB to Markdown - Upload a DRM-free .epub file in the EPUB to Markdown converter. The converter reads the whole book, so check the credit estimate for its length before you start. - Review the EPUB to Markdown output against the book. Check the chapter headings, a part or volume break, an illustration caption and the final chapter before reusing the text. - Copy or download the reviewed EPUB to Markdown result. Longer books and repeated conversions need a Lite, Pro or Max subscription; each plan has its own per-file page limit. ## Why use EPUB to Markdown - Chapters as headings — Books, parts and chapters become nested Markdown headings, so a long ebook keeps an outline you can navigate and split. - No contents clutter — The ebook's own linked contents page is removed, so the Markdown starts with the book instead of hundreds of links. - Captions kept apart — Illustration captions and meaningful alt text become separate lines instead of running into chapter titles. - Title and author up front — The book's title, author and language from its metadata sit at the top of the file for your records. - Credits known in advance — The estimate shows the page count and credits for the whole book before you convert, using the same count as the charge. - Browser and API — Check one book in the browser, then convert a reading list through the EPUB API route with the same results. ## Why ebook exports lose their chapters - Every heading looks the same — Many ebooks mark books, parts and chapters with the same heading tag, so a flat export cannot tell Book One from Chapter One. - The contents page repeats — The book's own table of contents becomes hundreds of links at the top of the text, before the first chapter begins. - Captions leak into titles — Illustrated editions put picture captions inside chapter headings, so a chapter title turns into a sentence from the picture. ## What this EPUB converter keeps - Chapter hierarchy — When the book's navigation nests chapters under parts, heading levels follow that nesting. When it is flat, a heading with no text of its own that is followed by chapters is treated as a part, and its chapters move one level down. - Contents pages and internal links — A contents page made of links to the book's own chapters is removed. Links to chapters that remain are kept; links to print page markers and other anchors that do not exist in Markdown become plain text. - Illustrations — Images are not embedded. A caption or meaningful alt text becomes its own line, such as *[Illustration: …]*, instead of running into a heading or paragraph. Cover images are left out. - Title, author and language — The book title and author from the package metadata appear at the top of the Markdown, with the language code. Translators and other contributors are not listed as authors. ## EPUB to Markdown in real workflows - Readers and researchers — Quote, annotate and search a public-domain book in a notes app, with chapter headings to cite. - AI reading assistants — Build a retrieval index from books so an assistant answers from the right chapter instead of guessing. - Authors and editors — Move a manuscript exported as EPUB back into Markdown for revision, version control or a static site. - Digital libraries — Turn a collection of DRM-free titles into plain text archives that stay readable without an ebook reader. ## EPUB to Markdown API Use POST /v1/convert/epub with an x-api-key header and a supported source file. A completed request can return outputs inline; HTTP 202 means conversion is still running. Read the Location response header and poll it before accessing file outputs. A multipart request accepts at most 100 files; subscription and total request limits also apply. Keep API keys in your server environment. API access requires a Lite, Pro or Max subscription. ```bash curl -i "https://api.markitdown.ai/v1/convert/epub" \ -H "x-api-key: $MARKITDOWN_API_KEY" \ -F "file=@source.epub" # If HTTP 202, GET the Location URL with the same API key. ``` ## Reading order and chapters EPUB to Markdown follows the book's spine, the reading order the publisher declared, rather than the order files happen to sit inside the archive. Each chapter file becomes consecutive Markdown, and the book's navigation decides how headings nest. That matters for long works: in War and Peace every Book and every Chapter uses the same heading tag, so a flat export shows fifteen books, two epilogues and three hundred and sixty-five chapters at one level. After conversion the books sit one level above their chapters, which is the outline a reader expects and the structure a splitter needs. Check one part break and the last chapter against your reading app before relying on the outline. ## Reviewing the output Review EPUB to Markdown output by comparing a few landmarks with the book in your usual reader. Open the first chapter, a chapter that starts a new part and the final chapter, then confirm the headings and their levels. Look at one illustrated page: its caption should appear as an illustration line near the text it accompanied, not inside a chapter title. Search the Markdown for a phrase from the middle of the book to be sure nothing was skipped. Front matter such as a publisher's licence or preface is kept, because it is part of the book; remove it yourself if your next step only needs the story text. ## Books for AI and search Converting an EPUB to Markdown gives a language model or a retrieval index text with its chapter boundaries intact. Split on chapter headings rather than on a fixed number of characters, and keep the book title and chapter name with every passage so an answer can point back to where it came from. A whole novel is far longer than most model context windows, so plan to retrieve relevant chapters instead of pasting the entire book into ChatGPT or Claude. The emphasis and quotation marks in the Markdown keep dialogue readable, which helps when you check a quoted passage against the original. ## Choosing the EPUB EPUB to Markdown needs a DRM-free file. Public-domain editions from Project Gutenberg or Standard Ebooks, your own manuscripts exported from a writing app and publisher review copies without copy protection all work. A book bought from a store that locks files to its reader app is encrypted and cannot be converted; the Markdown would be empty or rejected. If the same title is available as EPUB and PDF, choose the EPUB: its text and chapter structure are already there, while a PDF has to have its layout interpreted first. Use the latest edition you have, because corrected editions fix typos that conversion will otherwise faithfully copy. ## Credits for a whole book An EPUB has no fixed pages: text reflows to fit the reader's screen. EPUB to Markdown therefore bills by the length of the Markdown it produces, counting one page for every 1,800 characters at 1 credit per page. A full novel is long. Pride and Prejudice from Project Gutenberg converts to about 738,774 characters, which is 411 pages and 411 credits. War and Peace converts to about 3,222,019 characters, or 1,791 credits. The estimate shown before you convert uses the same count, so you see the number before any credits are spent. - Anonymous trial: one EPUB up to 10 MB and 25 pages, about 45,000 characters. That covers a short story or a few chapters, not a typical novel. - Lite: 150 credits a month and a 100-page limit per file, about 180,000 characters of Markdown. - Pro: 500 credits a month and a 200-page limit per file, about 360,000 characters. Pride and Prejudice, at 411 pages, is over this limit. - Max: 1200 credits a month and no per-file page limit, so a whole novel converts in one file as long as your balance covers it. War and Peace needs 1,791 credits, more than one month of Max; annual billing grants twelve months of credits up front. - A book over your plan's per-file limit is rejected before conversion and nothing is charged. Split it into smaller EPUB files, or upgrade. ## Example: War and Peace, Book One, Chapter I Source: Project Gutenberg #2600, War and Peace (EPUB), converted by markitdown.ai on 2026-09-28 ```markdown ## BOOK ONE: 1805 ### CHAPTER I “Well, Prince, so Genoa and Lucca are now just family estates of the Buonapartes. But I warn you, if you don’t tell me that this means war, if you still try to defend the infamies and horrors perpetrated by that Antichrist—I really believe he is Antichrist—I will have nothing more to do with you and you are no longer my friend, no longer my ‘faithful slave,’ as you call yourself! But how do you do? I see I have frightened you—sit down and tell me all the news.” It was in July, 1805, and the speaker was the well-known Anna Pávlovna Schérer, maid of honor and favorite of the Empress Márya Fëdorovna. With these words she greeted Prince Vasíli Kurágin, a man of high rank and importance, who was the first to arrive at her reception. Anna Pávlovna had had a cough for some days. She was, as she said, suffering from *la grippe; grippe* being then a new word in St. Petersburg, used only by the elite. ``` ## FAQ ### How does EPUB to Markdown work? EPUB to Markdown reads the chapters in the book's reading order and writes them as Markdown with headings, emphasis, lists and links. The anonymous trial accepts one EPUB up to 10 MB and 25 pages, with 3 conversions per IP per hour and 10 a day. ### How many credits does an EPUB book use? An EPUB book uses 1 credit for every 1,800 characters of Markdown it produces. Pride and Prejudice uses about 411 credits and War and Peace about 1,791. The estimate before conversion uses the same count. ### Which plan can convert a whole novel? Max can convert a whole novel in one file, because Max has no per-file page limit; Lite allows 100 pages per file and Pro 200. Your credit balance still has to cover the book. ### Are chapter and part headings kept? Yes. Chapter and part headings are kept and levelled from the book's navigation, so Book One can sit above Chapter I instead of beside it. ### What happens to images in an EPUB? Images in an EPUB are not embedded in the Markdown. Captions and meaningful alt text are kept as separate illustration lines; the cover image is left out. ### Can I convert an EPUB with DRM? No. An EPUB with DRM is encrypted and cannot be read. Convert a DRM-free EPUB, such as a public-domain title or a file you exported yourself. ### Is EPUB to MD the same as EPUB to Markdown? Yes. EPUB to MD means the same conversion: .md is the usual file extension for Markdown, and the download is a .md file. ### Do I need an account to try an EPUB? No. The anonymous trial lets you preview and copy the result of a short EPUB. A Lite, Pro or Max subscription is needed for longer books, saved history, batch and API access. ### Can I convert EPUB to Markdown with the API? Yes. POST the .epub to /v1/convert/epub with an API key; the estimate endpoint /v1/convert/epub/estimate returns pages and credits first. See /developers. ## Sources - [EPUB 3.3 (W3C Recommendation)](https://www.w3.org/TR/epub-33/) - [CommonMark Spec](https://spec.commonmark.org/) ## Related - [Markdown converter for documents and web pages.](https://markitdown.ai/) --- --- title: "Image to Markdown Converter" description: "Image to Markdown is a converter that reads PNG, JPG, WEBP, and TIFF with OCR today. The public trial accepts files up to 10 MB." url: https://markitdown.ai/image-to-markdown updated: 2026-09-26 source: markitdown.ai --- # Image to Markdown Converter > Image to Markdown is a converter that reads PNG, JPG, WEBP, and TIFF with OCR today. The public trial accepts files up to 10 MB. ## How to convert an image to Markdown - Upload a supported source in the Image to Markdown converter. Choose a file you can compare with the result, and check the anonymous trial limits before starting. - Review Image to Markdown output against the source. Check headings, table cells and the final passage before copying any content into another tool. - Copy the reviewed Image to Markdown output for your next task. For recurring account conversions, saved history, batch or API access, choose a Lite, Pro or Max subscription. ## Why use Image to Markdown - OCR for written content — Extract visible text from a screenshot, photograph or scan, then inspect uncertain characters before using the result. - A structured text result — Supported headings, paragraphs and tables become Markdown that is easier to review than an unstructured text dump. - No document wrapper needed — Start from the image you have instead of creating a Word document solely to paste a screenshot into it. - Check recognized numbers — Compare units, punctuation and numeric cells with the source image, especially when the result informs a decision. - Keep the visual source — Use the image alongside the Markdown when a chart, arrow or layout relationship cannot be explained by recognized labels alone. - The image API route — Send supported image files through the API after checking representative OCR outputs from the same source quality. ## Why a screenshot has no text layer - No text layer — Images carry no selectable text; the content has to be recognized before it can be used. - Screenshots pile up — Screenshots of documents are everywhere but hard to search until they become text. - Layout matters — Headings and tables inside an image should survive, not flatten into one line. ## What this image converter reads - Text extraction via OCR — Text is extracted from images through OCR today; accuracy varies by image quality, font clarity, and contrast. - Best for screenshots and scanned pages — Clean screenshots and scanned documents with clear print produce the best results; handwritten or low-resolution images may have lower fidelity. - Layout-heavy images — Complex layouts like multi-column magazine spreads or dense forms may not preserve spatial structure exactly, but the text content is still extracted. - Supported formats — PNG, JPG/JPEG, WEBP, and TIFF are all accepted, up to 10 MB on the public converter. ## Image to Markdown in real workflows - Screenshot notes — Recover written content from a screenshot and verify names, links and punctuation against the original. - Scanned records — Create editable text from a legible scan, keeping the image for any uncertain or consequential passage. - Research capture — Turn a photographed table into a draft text table and check its rows before analysis. - OCR evaluation — Compare results from clean and difficult source images before building repeated recognition into an application. ## Image to Markdown API Use POST /v1/convert/image with an x-api-key header and a supported source file. A completed request can return outputs inline; HTTP 202 means conversion is still running. Read the Location response header and poll it before accessing file outputs. A multipart request accepts at most 100 files; subscription and total request limits also apply. Keep API keys in your server environment. API access requires a Lite, Pro or Max subscription. ```bash curl -i "https://api.markitdown.ai/v1/convert/image" \ -H "x-api-key: $MARKITDOWN_API_KEY" \ -F "file=@source.png" # If HTTP 202, GET the Location URL with the same API key. ``` ## What OCR reads Image to Markdown uses OCR to recognize written content in an image and represent supported structure as Markdown. A clear screenshot of a document can produce useful headings, paragraphs and tables, while a blurry photograph may leave ambiguous characters. OCR reads visible text; it is not a guarantee that a chart’s trend, a diagram’s arrows or the meaning of a photograph will be explained. Distinguish recognized labels from a description of the visual itself. If the important information is only graphical, keep the image alongside the result and review it directly. Do not treat a short text export as proof that all of the source image’s meaning has been captured. ## Preparing the image Start image to Markdown conversion with the clearest original image you have. Repeatedly saving a compressed screenshot can blur punctuation and small text before OCR begins. Crop unrelated surroundings if you can do so without removing captions, units or table headings. Correct an obvious rotation and check that the source is readable at normal zoom. For a photograph, glare and perspective distortion can matter as much as resolution. If you cannot distinguish a character yourself, review the corresponding output carefully rather than expecting conversion to reconstruct missing evidence. Improving the source is usually more useful than sending the same damaged image repeatedly and choosing the most plausible-looking result. ## Reviewing the output Read image to Markdown tables row by row against the source, especially when the image contains dense numbers. A dropped decimal separator or a shifted value can change the meaning while leaving a table that still renders cleanly. Check column labels, units, totals and footnotes as separate pieces of evidence. If a table uses merged cells or visual groupings that Markdown cannot express directly, explain the grouping in the document where you use the result. Keep the screenshot available for a reader who needs to inspect the original. OCR saves retyping, but any consequential use of recognized numbers still needs a verification step. ## Image vs original Use image to Markdown when the image is the source available to you: a screenshot, a photographed page or an exported scan. If you also have the original Word document, spreadsheet or HTML page, choose its dedicated converter first. That source may expose structure and characters that an image has reduced to pixels. For a multi-page scanned report, the PDF converter is often the more natural place to keep the pages together. For a single screenshot, image OCR avoids creating an intermediate document merely to extract text. In either case, retain the source filename or reference in your notes so the Markdown can be traced back to the material you reviewed. ## OCR vs image understanding The public image converter is processed with OCR today: it reads printed or digital glyphs from the pixels and emits Markdown. That is not the same as AI image understanding, which paid-plan accounts can use at 5 credits per image and which produces an additional AI-enhanced Markdown output. This page does not treat understanding as the default path for a free screenshot. - OCR today is the shipped path for image uploads on this converter. - Paid-plan AI image understanding is billed separately at 5 credits per image. - Handwritten or low-contrast photos are a poor OCR fit; clean screenshots and scans work better. ## Scanned pages Scanned pages are a primary use case because they have no selectable text layer. Layout-heavy magazine spreads may not preserve spatial structure exactly, but the recognized text is still extracted. ## Example: OCR of the markitdown.ai pricing table Source: https://markitdown.ai/pricing ```markdown # Credits per page, by content. Text-layer documents and OCR pages cost the same per page. AI extras for figures and images are opt-in and metered separately. | Item | Best for | Rate | 150 credits buy | | --- | --- | --- | --- | | Documents | Text-layer PDF, Office files and HTML | 1 credit / page | 150 pages | | Scans & images | OCR for scanned pages, photos and screenshots | 1 credit / page | 150 pages | | AI figure enhancement | Charts and figures described in PDFs · paid plans | 3 credits / figure | 50 figures | | AI image understanding | Image content described, not just OCR · paid | 5 credits / image | 30 images | GOOD TO KNOW **Monthly reset** Monthly credits reset at the start of each billing period; unused credits don't roll over. **Annual up front** Annual plans are 20% off and grant the full year's credits at once — 1,800 on Lite. **One balance** The web app, Batch and the API all draw from the same credit balance. **When credits run out** Conversions pause until your next reset or an upgrade; upgrades top up the difference immediately. ``` ## FAQ ### How does Image to Markdown work? Image to Markdown converts the supported source into readable Markdown for review and reuse. The anonymous trial accepts up to 10 MB and 25 pages per file, with 3 conversions per IP per hour and 10 a day. Compare the result with the source before using it. ### What is the maximum image size? The public converter accepts image files up to 10 MB. ### How accurate is OCR on a screenshot? Images are processed with OCR today. Clean, high-contrast images produce better results than low-resolution or handwritten content. ### Is AI image understanding the default on this page? No. This converter uses OCR today. Paid-plan AI image understanding is a separate charge at 5 credits per image and is not the public default. ### Can it handle tables inside images? Table structure in images is recognized where possible; the fidelity depends on image clarity and how clearly the table grid is defined. ### Does it work on scanned documents? Yes. Scanned pages are a primary use case. Clear, high-resolution scans produce better Markdown output. ### What about handwritten content? Handwritten text is not reliably extracted by the current OCR pipeline. The tool works best with printed or digital text. ### Can I convert images from the API? Yes. Upload the image the same way you upload a PDF, then convert and poll. See /developers. ## Sources - [ISO/IEC 15948 PNG](https://www.w3.org/TR/png/) - [CommonMark Spec](https://spec.commonmark.org/) ## Related - [Markdown converter for documents and web pages.](https://markitdown.ai/) --- --- title: "HTML to Markdown Converter" description: "HTML to Markdown is a converter that turns uploaded .html files into structured Markdown for prompts. The public trial accepts files up to 10 MB." url: https://markitdown.ai/html-to-markdown updated: 2026-09-26 source: markitdown.ai --- # HTML to Markdown Converter > HTML to Markdown is a converter that turns uploaded .html files into structured Markdown for prompts. The public trial accepts files up to 10 MB. ## How to convert HTML to Markdown - Upload a supported source in the HTML to Markdown converter. Choose a file you can compare with the result, and check the anonymous trial limits before starting. - Review HTML to Markdown output against the source. Check headings, table cells and the final passage before copying any content into another tool. - Copy the reviewed HTML to Markdown output for your next task. For recurring account conversions, saved history, batch or API access, choose a Lite, Pro or Max subscription. ## Why use HTML to Markdown - Semantic headings — Convert meaningful HTML heading structure into a Markdown outline and review the resulting levels in the source view. - Links and lists — Carry supported link text and list structure into Markdown, then verify destinations in the context where you reuse the output. - Saved snapshot inputs — Process HTML you already have without fetching a live page that may have changed since the original export. - Code and table review — Check fenced code and simple tables in the destination renderer before publishing documentation migrated from HTML. - CSS stays out of the text — Focus on document content rather than reproducing positioning, animations or an interactive website theme. - Automation from HTML — Use the HTML API route for supported file or inline text inputs and handle asynchronous completion in your integration. ## Why HTML files still carry chrome - Boilerplate noise — Navigation, footers, and ads surround the content you actually want when the saved file includes the full page. - Nested markup — Deeply nested tags and inline styles obscure the real document structure. - Inconsistent headings — Heading levels are often misused in HTML; clean Markdown normalizes the hierarchy where tags are present. ## What this HTML converter keeps - File upload, not a fetch — This tool converts uploaded .html or .htm files. To convert a live web page by URL, use the URL to Markdown converter at /url-to-markdown. - Headings and tables — HTML heading tags (h1–h6) map to Markdown headings and table elements become Markdown pipe tables. - Inline scripts and styles — Inline script and style blocks are stripped from the output; only visible document content is kept. - Boilerplate in the saved file — The current converter preserves page text, which can include navigation, footers, and other boilerplate that was saved with the file. Review the Markdown before use. ## HTML to Markdown in real workflows - CMS exports — Prepare an exported article for a Markdown destination, checking links, media references and heading structure. - Documentation migration — Review code samples and tables in the new renderer before transferring a collection of HTML guides. - Archived snapshots — Work with a known saved page rather than depending on what a live URL happens to return today. - Text ingestion — Convert HTML collected by your own authorized workflow and review surrounding navigation before indexing it. ## HTML to Markdown API Use POST /v1/convert/html with an x-api-key header and a supported source file. A completed request can return outputs inline; HTTP 202 means conversion is still running. Read the Location response header and poll it before accessing file outputs. A multipart request accepts at most 100 files; subscription and total request limits also apply. Keep API keys in your server environment. API access requires a Lite, Pro or Max subscription. ```bash curl -i "https://api.markitdown.ai/v1/convert/html" \ -H "x-api-key: $MARKITDOWN_API_KEY" \ -F "file=@source.html" # If HTTP 202, GET the Location URL with the same API key. ``` ## HTML file vs URL HTML to Markdown starts with HTML you already possess. URL to Markdown starts with a network address and fetches the public page before conversion. Use a saved HTML file when you need to process a known snapshot, a template export or content that your own system has already collected. Use the URL tool when the public page is the source you want to retrieve now. The HTML converter does not execute scripts to fill in missing content, so a saved application shell may contain little of the article you saw in the browser. Inspect the source you submit and confirm that the text you need is actually present before troubleshooting the Markdown output. ## Reviewing the output HTML to Markdown can represent headings, paragraphs, lists, links, code and straightforward tables in a compact text form. CSS positioning and interactive controls have no direct equivalent in ordinary Markdown. A page that uses heading elements meaningfully gives the converter a clearer outline than a page made from styled generic containers. Review the exported heading hierarchy and a representative nested list in the Markdown source. For a table, confirm that the header still describes the values beneath it. The goal is a readable representation of the document’s content, not a reconstruction of the browser layout, hover behavior or visual theme. ## Navigation and links HTML to Markdown may retain navigation, footer text and other content alongside the article. Do not assume the converter has identified an article body or removed every advertisement. Review the beginning and end of the export for repeated menus before adding it to a retrieval index. A saved HTML file can also contain relative links whose meaning depends on the original site location. Check those destinations in the context where you publish the Markdown, and supply a meaningful source reference in your own workflow. Converting link syntax does not prove that a destination still exists, and a local file does not automatically provide the original website’s base address. ## CMS migrations Test HTML to Markdown with a representative post before exporting an entire CMS. Choose a post that includes a table, code, images, links and nested headings so you discover the important differences early. Compare the result in the destination renderer, because Markdown engines can differ in their support for extensions. Keep the original HTML export as an audit copy and record which article it belongs to. Decide in your migration process how to handle repeated navigation and media assets; this converter does not publish content or move a website’s files for you. A reviewed Markdown export is a portable starting point for that migration, not a completed site transfer. ## What the file keeps This tool converts a file you already have. It does not fetch the network. Navigation, footers, and other boilerplate that sit in the file can still appear in the Markdown — review before you embed. To convert a live web page by URL, use the URL to Markdown converter. - HTML heading tags (h1–h6) map to Markdown headings. - HTML tables become Markdown pipe tables. - Inline script and style blocks are stripped; only visible document content is kept. ## Example: markitdown.ai API docs Quickstart page Source: https://markitdown.ai/docs/quickstart ```markdown [Home](/) Getting started [Introduction](/docs)[Authentication](/docs/authentication)[Quickstart](/docs/quickstart) Core [Convert](/docs/convert)[Conversions](/docs/conversions)[Batches](/docs/batches)[Conversion Files](/docs/files)[Markdown Render](/docs/render) Developer [API keys](/docs/api-keys)[Webhooks](/docs/webhooks) Reference [URL fetcher](/docs/fetcher)[Errors & limits](/docs/errors-and-limits) API reference [All endpoints](/docs/api)[Error codes](/docs/errors) Quickstart Getting started # Quickstart Convert your first document to Markdown with the v1 API: one POST /v1/convert call, then read or poll the result. ``` ## FAQ ### How does HTML to Markdown work? HTML to Markdown converts the supported source into readable Markdown for review and reuse. The anonymous trial accepts up to 10 MB and 25 pages per file, with 3 conversions per IP per hour and 10 a day. Compare the result with the source before using it. ### What is the maximum HTML file size? The public converter accepts HTML files up to 10 MB. ### Is navigation and boilerplate removed from a saved page? The current converter preserves page text, which can include navigation, footers, and other boilerplate. Review the Markdown before use. ### How are HTML headings handled? HTML heading tags (h1–h6) map directly to Markdown heading levels, preserving the document hierarchy. ### Are HTML tables preserved? Yes. HTML tables are converted to Markdown pipe tables with their column structure intact. ### What about inline scripts and styles? Inline script and style content is stripped. Only the visible document text and structure is output. ### Is sign-up required to preview HTML? No. The anonymous trial lets you preview and copy a result. Account conversions and saved history require a Lite, Pro or Max subscription. ### Can I convert HTML from the API? Yes. POST the .html file to /v1/convert/html and poll the conversion if it does not finish inline. See /developers. ## Sources - [HTML Living Standard](https://html.spec.whatwg.org/) - [CommonMark Spec](https://spec.commonmark.org/) ## Related - [Markdown converter for documents and web pages.](https://markitdown.ai/) --- --- title: "URL to Markdown Converter" description: "URL to Markdown is a free converter that turns a publicly accessible web page into clean Markdown of its main content — the article, docs page, or thread without the site navigation — ready for LLMs, RAG pipelines, and notes. No sign-up to try; the trial allows 3 conversions per IP per hour and 10 a day." url: https://markitdown.ai/url-to-markdown updated: 2026-09-28 source: markitdown.ai --- # URL to Markdown Converter > URL to Markdown is a free converter that turns a publicly accessible web page into clean Markdown of its main content — the article, docs page, or thread without the site navigation — ready for LLMs, RAG pipelines, and notes. No sign-up to try; the trial allows 3 conversions per IP per hour and 10 a day. ## How to convert a web page to Markdown - Paste the address of a publicly accessible web page — an article, a docs page, a blog post — into the box above. - Press Convert. The server fetches the page, extracts the main content, and renders it in a browser first when the page needs JavaScript. - Read the result in Preview or Source, copy it, or sign in to download the .md file. ## Why use this URL to Markdown converter A web page is written for a browser. Markdown is written for people and models. This converter finds the part of the page that carries the meaning, keeps its structure, and leaves the site furniture behind. - Main content, not the whole page — Several extractors read the page and the result that keeps the most of the article wins, so menus, cookie banners, footers, and related-post rails stay out of the Markdown. - Structure that survives — Headings, lists, tables, and code blocks come through as Markdown, and every relative link and image path is rewritten to a full URL that still works outside the page. - JavaScript pages render too — When the first HTML is an empty application shell, the page is rendered in a headless browser and the rendered content is converted instead. - Free to try, no login — Preview and copy the Markdown without an account, up to 3 conversions an hour and 10 a day. Sign in to download the .md file. - A polite, identified fetcher — Every request identifies itself with a documented user agent, and site owners can block it — see the fetcher page in the docs. - Same engine as the API — The converter on this page and POST /v1/convert/url run the same code, so a script can convert up to 20 URLs per call. ## Why a live page is noisy for models - Pages are noisy — Live web pages bury content under navigation, ads, banners, and scripts — often several times more text than the article itself. - Markup is messy — Hand-written and generated HTML rarely maps cleanly to readable structure. - AI needs the article — RAG and agents want the article, not the page chrome around it — and some pages only show the article after JavaScript runs. ## What this URL converter fetches - Server-side fetch — The page is fetched server-side, asking for Markdown first and HTML second; a site's own Markdown is used only when it passes a quality check. No browser plugin or local tool is needed. - Main-content extraction — Several extractors — readability scoring, boilerplate detection, text-density analysis, and the page's own main or article element — compete, and the one that keeps the most structure and prose wins. Well-known sites such as documentation generators, GitHub, and Wikipedia use site rules. - JavaScript-rendered pages — An empty application shell is rendered in a headless browser, which waits for the page to settle and then hands the rendered DOM to the same extraction. Content that appears only after a click or a scroll is not captured. - Public pages, size and identity — Only publicly accessible URLs are supported: pages behind authentication, paywalls, or login walls cannot be fetched. Fetched pages are subject to response-size limits, and the fetcher user agent is documented on /docs/fetcher. ## Use cases for website to Markdown conversion From AI agents to content migration, converting a webpage to Markdown is a small step that many workflows depend on. - AI and LLM pipelines — Feed articles and documentation into RAG indexes, agent tools, and prompts as clean text with its headings intact, instead of raw HTML. - Research and notes — Save a source as Markdown to quote, annotate, or keep in Obsidian or a notes repository — with the original links still in place. - Content teams — Move posts and help-center pages from one CMS to another, or pull reference material into a Markdown-based docs site. - Developers — Call the API from ingestion jobs and crawlers you already run, and get the same Markdown the browser converter shows. ## URL to Markdown API for developers Converting pages from code? Send one URL — or up to 20 in a urls array — to POST /v1/convert/url with your API key. Set render to auto, force, or skip to control JavaScript rendering, and frontmatter to true to put the page title, author, and publish date in YAML front matter. The Markdown comes back in the same response when the page finishes within the wait window; otherwise you get a conversion to poll, or a webhook when it settles. API access is on paid plans. ```bash curl -X POST https://api.markitdown.ai/v1/convert/url \ -H "x-api-key: mdai_…" \ -H "content-type: application/json" \ -d '{"url":"https://markitdown.ai/blog/why-pdfs-break-llms","frontmatter":true}' ``` ## Reviewing the output Review URL to Markdown output beside the public page you submitted. The converter keeps the main content and leaves out navigation, cookie notices, and footer links, but a page without a clear article — a product page, a landing page, a home page with many teasers — can still lose a side panel or keep a block you would not quote. Check the first and last paragraphs against the page before you index the result. Pages that build their content with JavaScript are rendered in a browser first; a page that loads its text only after a click, a scroll, or a login still cannot be captured. When a site already serves its own Markdown to clients that ask for it, the converter uses that version if it passes a quality check, and falls back to the HTML page when it looks like a whole-site menu or an archive dump. Keep the source URL with the Markdown so a reader can return to the original page. ## Keeping source context Keep the submitted URL and your retrieval date with a URL to Markdown result in your own workflow. A public page can change after you convert it, so a later visitor may see a revised passage. With front matter turned on, the result opens with the source URL, the canonical URL, the title, and — when the page declares them — the author, publish date, description, site name, language, and lead image, read from the page's meta tags, structured data, and byline. Those fields repeat what the page says about itself; check them at the source if your research depends on them. The API metadata field is different: it only carries context you supply. Before quoting a table or paragraph, compare it with the page and retain enough surrounding context to avoid changing its meaning. Markdown makes web content easier to reuse; it does not establish whether the page is accurate or whether you have permission to republish it. ## Saved HTML vs URL Choose the HTML converter when you already possess the exact page snapshot you need to process. Choose URL to Markdown when a publicly accessible address is the input for the current conversion. These are related workflows with different source controls: a saved file fixes the input you submit and is converted in full, while a URL depends on what the website returns at the time of the request and is reduced to its main content. Neither route bypasses access restrictions. If a site denies the fetcher or requires a login, use material you are authorized to obtain through an appropriate workflow. Do not repeatedly retry an inaccessible URL as though another conversion attempt grants access. ## LLM research Researchers paste a source, not a file. This converter fetches the URL you submit, keeps the article or documentation body, and returns Markdown you can quote, chunk, or drop into a notes vault, with links that still resolve. It renders JavaScript pages when the server sends an empty shell, and it does not log into paywalls. - Only publicly accessible URLs are supported. - The fetcher identifies itself; operators can deny the user agent — see /docs/fetcher. - Saved HTML files belong on /html-to-markdown, not here. ## Example: markitdown.ai blog post via URL fetch Source: https://markitdown.ai/blog/why-pdfs-break-llms ```markdown # Why PDFs break LLMs — and what clean Markdown fixes _Published 2026-06-12 · The markitdown.ai team_ > A PDF that looks perfect to a human often falls apart the moment a model reads it. Here's what goes wrong, and why converting to clean Markdown first makes downstream AI far more reliable. A PDF renders perfectly for a human and falls apart the moment a model reads it. The page you see — columns, tables, headers, footnotes — is a *visual* arrangement, not a logical one. When you pull raw text out of it, that visual order rarely survives, and whatever you feed a language model is only as good as the text it actually receives. ## What actually goes wrong When teams pipe documents straight into an LLM or a retrieval index, a few failure modes show up again and again: - **Reading order breaks.** Multi-column layouts interleave. A two-column page becomes "line 1 left, line 1 right, line 2 left, line 2 right…", so a sentence a human reads top-to-bottom arrives at the model scrambled. - **Tables collapse.** Copy-paste turns a table into a loose run of numbers. The relationship between a header and its cell — the entire point of a table — is gone. - **Headings disappear.** The visual hierarchy (this is a section, this is a subsection) is encoded as font size and weight, not structure. Strip the styling and you get one undifferentiated wall of text. ``` ## FAQ ### How does URL to Markdown work? URL to Markdown fetches the public page you paste, keeps its main content, and converts that to Markdown with full-URL links — no file to upload. Pages that need JavaScript are rendered in a browser first. ### Does it strip ads and navigation from a live page? Yes, for pages with a clear main content area. URL to Markdown keeps the article, documentation body, or thread and drops menus, cookie banners, and footers; a landing page or product page without one main block can still keep or lose a side panel, so check the result. ### Are there size limits for a fetched page? The public converter works on public web pages. Fetched pages are subject to response-size limits and can be rejected if too large. ### Does it work on pages behind a login? No. Only publicly accessible URLs are supported. Pages behind authentication, paywalls, or login walls cannot be fetched. ### What about JavaScript-rendered pages? JavaScript-rendered pages are rendered in a headless browser when the first HTML is an empty shell, then converted like any other page. Through the API, render set to force always renders and skip never does. Text that appears only after a click, a scroll, or a login is not captured. ### Can I convert multiple URLs at once? Send a urls array with up to 20 public URLs to POST /v1/convert/url. Track each file result and handle asynchronous responses. See /developers. ### Is URL to Markdown free? Yes. Previewing and copying the Markdown is free without an account, up to 3 conversions an hour and 10 a day; downloading the .md file needs a sign-in. Public web pages only (no internal or private URLs). ### What content does the converter keep? The converter keeps the main content's headings, paragraphs, lists, links, images, tables, and code blocks, with relative links rewritten to full URLs. Scripts, styles, navigation, and footers are dropped; with front matter on, the page title, author, and publish date go at the top. ### Is there a URL conversion API? Yes. Convert web pages by URL programmatically via the REST API. See /developers. Saved HTML files use /html-to-markdown instead. ## Sources - [HTML Living Standard](https://html.spec.whatwg.org/) - [RFC 9110: HTTP Semantics](https://www.rfc-editor.org/rfc/rfc9110) ## Related - [Markdown converter for documents and web pages.](https://markitdown.ai/) --- --- title: "Markdown to HTML Converter" description: "Markdown to HTML is a browser converter that turns CommonMark and GitHub Flavored Markdown into sanitized HTML — headings, tables, task lists and fenced code included. Nothing is uploaded; the same renderer runs behind POST /v1/convert/markdown/html." url: https://markitdown.ai/markdown-to-html updated: 2026-09-26 source: markitdown.ai --- # Markdown to HTML Converter > Markdown to HTML is a browser converter that turns CommonMark and GitHub Flavored Markdown into sanitized HTML — headings, tables, task lists and fenced code included. Nothing is uploaded; the same renderer runs behind POST /v1/convert/markdown/html. ## How to convert Markdown to HTML - Paste or type Markdown into the editor. The preview renders locally in your browser as you type — no file leaves your device. - Review the HTML output against your Markdown. Check headings, tables, task lists and code blocks, then choose a fragment or a full document. - Copy the HTML or download the file for your page or CMS. Automate the same output with POST /v1/convert/markdown/html and your API key. ## Why use this Markdown to HTML converter A CMS field, a static site and an email editor all want markup, not Markdown source. This converter keeps the parts that carry meaning and strips what would make raw HTML unsafe. - Runs in your browser — Paste Markdown and the HTML preview updates locally. Nothing is uploaded, and no account is required to preview, copy or download. - CommonMark plus GFM — Headings, lists, links, blockquotes and fenced code follow CommonMark; tables, task lists and strikethrough follow GitHub Flavored Markdown. - Sanitized by default — Script tags, inline event handlers and unsafe href or src schemes are stripped, so the HTML is safe to paste into a page you don't fully control. - Fragment or full document — Copy just the rendered markup for a page that already owns its layout, or download a complete HTML document with the previewed styling. - Heading anchors — Each heading gets a stable, URL-safe id generated the same way the API does, so a table of contents or a deep link keeps working after you publish. - Same engine as the API — This preview and POST /v1/convert/markdown/html run the same renderer, so the HTML you copy here matches an automated call byte-for-byte. ## Why raw Markdown can't go straight into a web page - Markdown syntax is not HTML — A CMS field, an email template or a static-site theme wants markup, not Markdown source, so the raw text has to be converted before it can be styled or embedded. - A plain parser misses GFM — Tables, task lists and strikethrough are GitHub Flavored Markdown extensions; a CommonMark-only parser renders them as literal pipe characters and tildes instead of structure. - Untrusted Markdown can carry scripts — Markdown can embed raw HTML, including