Image to Markdown Converter

Convert Image to Markdown with private in-browser OCR. Turn screenshots, scans, PNG, JPG, WebP, and GIF files into editable headings, lists, and tables without an upload.

PNG to Markdown: improve the source

Image to Markdown starts with OCR. Recognition quality depends on the image, so improving the source often helps more than editing the generated Markdown afterward.

The OCR engine and selected language data are downloaded when needed; your image stays on your device. Recognition does not establish factual accuracy. Check identifiers, amounts and dates against the original before publishing, and retain the source image when exact wording matters.

Screenshot with small text
Capture at the original display resolution. Crop away browser chrome and unrelated panels, while keeping the full text lines and the table’s outside edges.
Phone photo of a page
Place the page flat, keep the camera parallel, avoid shadows and glare, and rotate the image upright. Perspective distortion makes character recognition harder.
Table or code screenshot
Compare columns, indentation and punctuation carefully. OCR can confuse 0/O, 1/l and paired symbols. It cannot reliably infer every merged cell or reconstruct code semantics.
Scanned PDF or a text document
Use PDF to Markdown for multi-page PDFs. If you have the original Word or HTML file, convert that instead of a screenshot to preserve more structure.

Maintained by Markdown Viewer. Reference: Tesseract: improving recognition quality. Report a conversion issue.

What Is the Image to Markdown Converter?

The Image to Markdown converter reads the text in a screenshot, scan, or photo and writes it back out as Markdown — headings, lists, and tables included — using OCR that runs inside your own browser rather than on a server.

  • Structure, not just characters: Type size becomes headings, bullets become list items, and aligned words become GFM tables.
  • Paste, drop, or open a batch: Paste or add up to twelve images; one engine reads the queue and exports the finished Markdown as individual files or a ZIP.
  • Languages and confidence: Choose English or a mixed-script language pair, then use the reported confidence to decide what needs proofreading.
The image to markdown converter with a scanned page on the left and the recognised Markdown rendered on the right, above a confidence readout.
A scanned page, the reconstructed Markdown, and the confidence the engine reports for it.

How Image to Markdown Conversion Works

Image to Markdown recognises words and positions, rebuilds document structure, and keeps the result adjustable.

  1. 1The engine loads once, in your tabTesseract's WebAssembly core and the language data you picked are fetched the first time an image lands, then cached by the browser. The whole queue is read with that one load.
  2. 2Recognition returns words and boxesEach word comes back with its text, its confidence, and the rectangle it occupies. Low-confidence fragments are dropped rather than guessed at, and counted for you.
  3. 3Geometry becomes structureLine heights are compared against the body text to find headings, leading glyphs and indents to find lists, and repeated x positions across rows to find table columns.
  4. 4Markdown is emitted, and stays adjustableThe layout is kept in memory, so toggling heading, list, or table detection re-derives the Markdown in milliseconds instead of reading the image again.
What the OCR pass seesin
"Follow-up actions"   18pt bold   y=557
"Item"  "Owner"  "Due"    11pt       y=599
"Clear drainage outlet"  …  x=372, 723, 938
What the converter writesout
## Follow-up actions

| Item | Owner | Due |
| --- | --- | --- |
| Clear drainage outlet | Maintenance | 14 days |

An Image to Markdown Tool That Keeps the Structure

Image to Markdown uses OCR geometry to preserve headings, lists, tables, languages, and readable paragraphs.

Headings From Type Size

Text noticeably larger than the body of the page becomes a # heading at the matching level. Size ratio drives the decision, so the same rules work on a phone photo and a 600 DPI scan.

Tables Rebuilt From Column Positions

Rows whose words land in the same vertical bands are rebuilt as a GFM pipe table — the hardest part of Image to Markdown conversion, and the one worth checking.

Lists and Paragraphs, Not Line Soup

Bullet glyphs and 1. / a) markers become nested list items. Lines broken by the page margin are joined back into paragraphs, words hyphenated across a break are rejoined, and a wide vertical gap starts a new block.

Multilingual Recognition

English plus Chinese, Japanese, Korean, Russian, Arabic and the major European languages, in mixed-script pairs so a bilingual document is read in one pass. Switching the language reads every open image again in it.

Confidence You Can See

The converter reports its own average confidence and how many uncertain words it discarded, so you know whether to trust the result or rescan at a higher resolution.

Private Image to Markdown

Recognition runs locally on your own machine. Unlike hosted OCR services, your screenshot, contract scan, or whiteboard photo is never transmitted to or stored on a server.

Convert an Image to Markdown in 3 Steps

Convert Image to Markdown from a screenshot, scan, or photo without an account or upload.

1

Add one image or twelve

Paste a screenshot or add PNG, JPG, WebP, and GIF files up to 15 MB each; every image joins the conversion queue.

2

Watch it read them

Image to Markdown starts automatically and reads the queue with one shared OCR engine. Choose the language before or after adding images.

3

Check the structure, then copy

Compare the preview with the source, adjust heading, list, and table detection, then copy Markdown, download .md, or export the batch as a ZIP.

Image to Markdown Questions

Answers about OCR accuracy, privacy, languages, tables, code, handwriting, and caching.

How do I convert an image to Markdown?
Paste a screenshot with Ctrl+V, or drop one or more image files onto the page. That is the whole instruction — conversion starts on its own, with no Convert button to press. The text is recognised in your browser and rebuilt as Markdown with headings, lists, and tables inferred from the layout, shown as a rendered preview next to the picture. Copy the result, download it as a .md file, or download a batch as a ZIP.
Is my image uploaded anywhere?
No. The recognition engine runs inside your browser, so the picture itself is never transmitted, queued, or stored on a server. Be aware of one detail we would rather state plainly: the engine and its language data are downloaded from a public CDN the first time you convert, and cached by your browser afterwards. That download contains no part of your image.
How accurate is the OCR?
On clean, high-resolution screenshots of printed or rendered text, accuracy is high — often near perfect for body text. On photographs, low-resolution captures, skewed pages, unusual fonts, or handwriting it drops sharply. The converter reports its average confidence so you are not guessing. Always proofread the output before relying on it.
Can it read handwriting?
Not reliably. The engine is built for printed and rendered text. Neat block capitals sometimes come through; ordinary cursive generally does not. For handwritten notes, expect to do substantial correction, and treat the result as a starting point rather than a transcript.
Does it convert tables in an image to Markdown?
Yes, by looking at geometry rather than gridlines. When at least three consecutive lines split into the same number of cells at the same horizontal positions, they are rebuilt as a GFM pipe table. Dense tables, merged headings, and columns separated by very little whitespace are where this is most likely to need a fix by hand — check the result against the image.
Can it extract code from a screenshot?
The text comes through, but not as a fenced code block, and this is a genuine limitation worth knowing about. OCR frequently confuses characters that matter enormously in code — l against 1, O against 0, curly against straight quotes — and it does not preserve exact indentation. Use it to avoid retyping, then read the code carefully before you run it.
Are mathematical formulas converted to LaTeX?
No. Recognising a formula's two-dimensional structure and re-expressing it as LaTeX is a different class of model from text OCR, and this converter does not attempt it. A formula will come out as a rough, usually incorrect line of characters. For equations, a dedicated maths-recognition tool is the right choice.
Which languages are supported?
English on its own, or English paired with Simplified Chinese, Traditional Chinese, Japanese, Korean, French, German, Spanish, Portuguese, Russian, Italian, or Arabic. Pairing means a bilingual page is read in a single pass. Each extra language adds to the one-time download.
Why is the first conversion slow?
The recognition engine and language model are fetched on first use, then cached by your browser. Subsequent conversions start faster and work offline with languages that have already been cached.
How can I get better results?
Give it more pixels: a 2x screenshot beats a downscaled one. Crop to just the region you need, straighten a skewed photo before converting, prefer flat even lighting over a shadow across the page, and pick the right language. Dark-mode screenshots of light text usually work fine.
Can I batch convert many images?
Up to twelve at a time. Drop or paste several images and they queue up, each becoming a row in the list on the left with its own Markdown, and one engine load reads the whole queue — including anything you drop while it is still working. Switch between the images to check each result, then take them all as a ZIP. For a large archive of scans, a server-side OCR pipeline or an API-based service still gives you the throughput, retries, and logging that a browser tab cannot.

Convert an Image to Markdown Now

Paste a screenshot or add an image, then turn unselectable text into editable Markdown.

Last updated