From Screenshot to Markdown: Why I Built an Image to Markdown Tool

A short build-in-public note about my thinking about the TO MD product.

Hello DEV community!

I recently built a small Markdown conversion project called TO MD.

The first problem I wanted to solve was simple:

When I see some great words, I'd like to snapshot them and the information is very useful, but they're trapped trapped inside screenshots .

Screenshots are easy to share, but hard to reuse. A screenshot may contain a table, a receipt, a document section, or a few lines of important context. It looks readable, but the content is not searchable, editable, or easy to paste into docs.

So I started building an image to markdown workflow.

The goal was not just OCR. I wanted a Markdown draft that preserved a useful structure.

Why screenshots are awkward for documentation

Markdown is portable.

It works in READMEs, GitHub issues, docs sites, internal notes, and AI prompts. You can edit it, diff it, search it, and reuse it.

Images are different.

Once information is inside a PNG or JPG, a few things become harder:

• copying the text
• reusing a table
• searching the content
• cleaning up a document
• giving an AI assistant structured context I kept running into this when working with screenshots of documents and table-like content. Manually rewriting them into Markdown felt like wasted time.

What I wanted the output to look like

A basic OCR result is useful, but often not enough.

If the source image contains a document, the Markdown should try to keep the shape of that document:

• headings should remain readable
• line breaks should not disappear
• table-like areas should become Markdown tables when possible
• the result should be easy to copy or download as .md

For example, if an image contains transaction rows, I do not want one long paragraph. I want a table draft that I can review and fix.

That review step matters. OCR is not perfect, especially with small numbers, dates, currency symbols, rotated photos, or compressed JPG files.

But even an imperfect structured draft is much faster than starting from zero.

Why Markdown is still a good target format
Markdown is lightweight, but structured enough.
That makes it a useful bridge between visual information and reusable text. Once an image becomes Markdown, it can move into:

• documentation
• GitHub issues
• project notes
• static sites
• AI workflows
• internal knowledge bases

This became more important to me as AI tools became part of daily work. A clean Markdown version of a screenshot is often much easier to use as context than the original image.
What I learned while building it

1. Structure matters more than raw text

The hard part is not extracting some text.

The hard part is returning text that still feels useful. Headings, tables, spacing, and line breaks make the difference between “technically extracted” and “ready to edit.”

2. Source image quality matters a lot

PNG screenshots often work well because text edges stay sharp.

JPG can work too, but compression may blur small text or table lines. A high-resolution original image usually produces a better Markdown draft than a resized or forwarded chat image.

3. The user still needs control

I do not think tools like this should pretend to be perfect.Therefore, manual audit is always necessay. Therefore, i prepared a live preview for the markdown. Now the workflow works like the following:

1. upload the image
2. convert the image into Markdown
3. review the result
4. fix anything important
5. copy or download the final .md

That is especially true for financial, legal, medical, or internal business documents.
What I shipped first

The first version focuses on practical conversion workflows:

• image to Markdown
• PDF to Markdown
• Word to Markdown
• HTML to Markdown
• text to Markdown
• JSON to Markdown
• Markdown to PDF

For images, the main use case is turning screenshots, scans, and document images into editable Markdown drafts.
I am still improving table reconstruction, document layout handling, and examples for different image types.
What I am thinking about next

Some areas I want to improve:

• better Markdown tables from screenshots
• support for more complex document layouts
• cleaner handling of receipts and invoices
• batch image conversion
• more before-and-after examples
• better guidance for reviewing OCR output

The deeper product question is:
How do you turn visual information into text without losing the structure that made the information useful?
That is what I am trying to solve.

Feedback welcome
If you work with documentation, screenshots, AI prompts, or scanned files, I would love feedback.
What kinds of images do you most often need to turn into structured text?
And what functions do you suggest for the daily working?