Turn any image into clean Markdown.
Screenshots get OCR, photos get an AI description, mixed images get both. The converter decides automatically.
Your files are never stored: documents convert in your browser where possible, and anything sent to our AI travels encrypted (HTTPS) and is discarded after processing. Privacy Policy
Not every image should be treated the same way. CleanMD classifies each image first: a screenshot of text goes through OCR and comes back as structured Markdown; a photo or illustration gets a faithful AI description; a slide with both text and graphics gets a combination. You never have to pick a mode — the routing is automatic.
Working with related formats? See also PDF to Markdown, PPTX to Markdown.
Text, picture, or both
An image can be three different things. A photo of a printed page is really a document, and what you want back is the text with its headings and tables. A photo of a whiteboard sketch or a product has no text worth extracting, so the useful output is a description. Slides, infographics and screenshots of dashboards are a mix, with text that matters and a layout around it that carries meaning too.
CleanMD sends the image to an AI model once and lets it decide which of the three it's looking at. For a page it transcribes the text and rebuilds the hierarchy, so a heading on the photographed page becomes a Markdown heading. For a picture it writes a description. For the mixed case you get the extracted text first and then a description of what's around it.
Parts that can't be read are marked as unreadable instead of being filled in. We'd rather leave a gap than invent a number, and this matters most for photos taken at an angle or in bad light, where the bottom lines of a page are often too blurred to trust.
You don't need to save a screenshot first. Paste it with Cmd+V or Ctrl+V on the home page and it goes straight to the converter, which is the quickest way to turn a screenshot of a table or an error message into text you can edit.
JPG, PNG, WEBP and GIF files up to 3 MB are accepted, and image conversion is part of the free tier. Unlike documents, images can't be read in the browser, so they're sent to the model and deleted when the request ends. If you have many pages of the same document, a single scanned PDF usually works better than separate photos, because the OCR engine then sees the pages in order and keeps the headings consistent from one page to the next.
How it works
- 1Drop a JPG, PNG, WebP or GIF in the converter above.
- 2The AI classifies the image (text, visual, or mixed) and routes it to OCR, description, or both.
- 3Review the Markdown and download the .md file.
Before and after
[Screenshot of a pricing table] Starter $9/mo 1 project Pro $29/mo 10 projects Team $99/mo Unlimited
## Pricing | Plan | Price | Projects | | ------- | ------ | --------- | | Starter | $9/mo | 1 | | Pro | $29/mo | 10 | | Team | $99/mo | Unlimited |
Frequently asked questions
How does it decide between OCR and description?
Each image is classified by an AI model first: mostly text → OCR, mostly visual → description, mixed → both combined. You don't have to choose anything.
Which image formats are supported?
JPG, JPEG, PNG, WebP and GIF, up to 3 MB on the free tier.
Can it read tables inside screenshots?
Yes. Tables detected in the image are reconstructed as real Markdown tables rather than lines of loose text.
Is the image stored after conversion?
No. Images are processed only for the duration of the conversion and nothing is persisted afterwards.