Workers AI's toMarkdown can now read GIFs and BMPs too!

Hi everyone, it's Shiichan! Today I found a small but genuinely useful Workers AI update I wanted to share.

Cloudflare Changelog developers.cloudflare.com

What was announced?

From Cloudflare's Changelog: Workers AI's Markdown conversion feature, toMarkdown, now supports GIF and BMP image files. It already handled JPEG, PNG, WebP, and SVG, and now GIF and BMP join that list. AI Search, which indexes your files for search, uses toMarkdown automatically, so GIF and BMP files get the same benefit without you having to do anything extra.

The story so far

Until now, toMarkdown covered four formats: JPEG, PNG, WebP, and SVG, but not GIF or BMP. That meant if your knowledge base had a mix of screenshots and older GIF or BMP scans, those files simply couldn't be converted, and slipped out of search results.

What changes

GIF and BMP files now run through the same processing pipeline as the other image formats and get converted into Markdown. If you use AI Search, your GIF and BMP files will be picked up automatically the next time your index syncs, and there's nothing you need to reconfigure.

Dive Deep

According to the post, images go through this pipeline:

• Images are resized if needed
• For animated GIFs, only the first frame is used
• An object-detection model identifies what's in the image
• A vision model then writes a natural-language description based on the detected objects
• That description becomes searchable, machine-readable Markdown

This is the exact same pipeline used for JPEG and PNG, so GIF and BMP aren't treated as special cases. It should quietly help in places where formats tend to mix, like a support knowledge base full of screenshots or an archive of old BMP scans. If you want the details, check the Markdown conversion documentation and the list of file types AI Search supports.

Wrap-up

• Workers AI's toMarkdown now converts GIF and BMP images
• For animated GIFs, only the first frame gets processed
• It runs through the same pipeline as other formats: object detection followed by a vision model writing the description
• AI Search will pick up GIF and BMP files automatically on your next index sync, with no configuration changes needed

If you're using AI Search over a knowledge base with a mix of screenshots and old scanned images, this is a small but welcome update!