How to Turn a Screenshot Into Text? In-Browser OCR That Never Uploads Your Image
Upload-based OCR versus TwoThink's local image recognition: when to use each, how it differs from PDF text extraction, and how to handle language packs and proofreading.
A chat screenshot, a photo of a whiteboard, a scanned page saved as JPG — you want the text, but you do not want the image on a stranger's server. TwoThink Image OCR runs tesseract.js inside your browser, and the file is never uploaded.
Which one: this or PDF text extraction?
| Extract PDF Text (PDF site) | Image OCR (this site) | |
|---|---|---|
| Input | PDF files | JPG / PNG / WebP |
| How it works | Reads the text layer first, OCR only for scans | Always OCR |
| Typical use | Contracts, papers | WeChat screenshots, photographed documents |
| Language packs | /tesseract/ on the PDF site | /tesseract/ on this site (each site is independent) |
If you have a PDF, use the PDF site; if you only have an image or a screenshot, use the OCR here.
How to run it
- Open Image OCR
- Drop in one image (up to 50MB)
- Choose the language (Simplified Chinese and English by default) → Start recognition
- The first run loads the language pack (about 10–15MB, served from this site's own static assets); the browser can cache it afterwards
- Proofread in the result box → Copy or Download TXT
You can hit Cancel to stop the worker while it is running.
What accuracy to expect
- Printed text and clean screenshots: usually usable
- Handwriting, tilted pages, blur, busy backgrounds: proofread by hand
- It does not rebuild tables, analyse layouts, or translate (the same limits as the PDF OCR)
Privacy and the network tab
The OCR engine, the WASM runtime and the language packs all load same-origin from img.twothink.xyz; you should not see requests to tesseract.projectnaptha.com or any other overseas CDN. The image data is only handled inside a local worker.
If you are on WeChat
Large WASM plus a language pack is unstable in the WeChat in-app browser. Please use ··· → Open in browser. Lighter tools such as compression and cropping still work inside WeChat.
Related tools
- Cut out first, then lay it out: Remove Image Background → ID Photo Generator
- Sharing a long image: Stitch Images Into a Long Image
- Text out of a PDF: Extract PDF Text
Entirely local, and no sign-up.