HOW IT WORKS
Three clicks, start to finish.
Drag a box over anything
Click Capture region and draw a rectangle over any part of the screen. PixText grabs exactly what you selected — no fiddling with file exports.
AI reads the pixels
The cropped image goes to Google's Gemini vision model, which reads messy, low-res, and handwritten text the old Tesseract clones choke on.
Copy, overlay, or keep it
Copy the text, or hit Locate to draw it back over the page in place. Every capture is saved to a searchable history with its thumbnail.
WHY THIS ONE
Built to beat the incumbents where they’re weak.
Vision-LLM accuracy
Reads low-res, skewed, and handwritten text that local Tesseract engines miss — because it's an actual vision model, not pattern matching.
Overlay back on the page
The only OCR tool that draws the recognized text back over the original, aligned to the page and staying put as you scroll. Click any box to copy just that line.
Searchable history
Every capture is kept locally with its thumbnail so you can find that snippet again. Nothing leaves your device except the image you're reading.
Your image, your key
Unlike cloud OCR services, your screenshot goes only to Google's Gemini API — with your own key, or our proxy. Never a third-party OCR server.
FREE, OR ZERO-SETUP
Two honest doors — pick the one that fits you.
This is an AI tool, so it needs a Google Gemini model to run. Bring your own key and it’s free forever — your data goes straight to Google, never through us. Or subscribe once and skip all setup, here and in every other Browserbelt tool.
| FREE FOREVER Bring your own key | $5/MO · $39/YR Browserbelt subscription | |
|---|---|---|
| Setup | Paste your Gemini API key once | None — sign in and go |
| Who pays for AI | You pay Google directly, at cost — we never mark it up or see your key | We do; one flat, predictable price |
| Data path | Your browser → Google's API. Nothing through our servers | Through our proxy (required to meter usage), straight to Gemini, never stored |
| Limits | Whatever your Google quota allows | Fair-use cap (~1,000 AI actions/mo) |
| Best for | Tinkerers who already have an API key | Everyone who just wants it to work |
FAQ
Questions, answered plainly.
How is this better than Copyfish or a Tesseract extension?
Copyfish sends your image to its own cloud OCR service; the Tesseract clones run locally but are mediocre on messy or low-res text. PixText uses a vision LLM for accuracy and sends your image only to Google's Gemini — with your own key it never touches a third-party server at all.
What exactly gets sent, and where?
Only the rectangle you drag — as an image — goes to Gemini for reading. The rest of the page is never captured. Your OCR history and API key stay in local storage on your device.
Do I have to pay?
No. This tool is free forever with your own Google Gemini API key — paste it once and your data goes straight from your browser to Google, never through us. If you'd rather skip the setup, one Browserbelt subscription ($5/mo or $39/yr) turns on instant AI here and in every other Browserbelt tool, with no key to manage.
Does it work on chrome:// pages and the Web Store?
No — Chrome blocks extensions from reading those protected pages, so Capture is disabled there with a note. Everywhere else works.
Full data details: PixText privacy policy.
MORE FROM BROWSERBELT
