Why PasteOCR
Paste and go
The whole page listens for Ctrl+V. Pasting a screenshot starts recognition instantly — no buttons to click.
Fully local AI engine
Official PaddleOCR models run in your browser via WebAssembly. Mixed Chinese–English text just works, with no language to pick.
Privacy first
Images never leave the browser: no uploads, no analytics scripts, no ads. Safe even for sensitive screenshots.
Structured extraction
Domains, URLs, emails and long digit strings are extracted automatically — copy them one by one or all at once.
Download once, instant after
The model (~6MB) is downloaded once and cached. Later visits start instantly and typical screenshots finish within a few hundred milliseconds.
Free without limits
No install, no sign-up, no usage caps, no watermark. Works in desktop and mobile browsers.
Recognize text in three steps
-
Paste or drop an image
Press Ctrl+V, drag an image into the dashed box, or click to choose a file.
-
Local recognition runs
The model downloads automatically on first use; from then on, recognition starts immediately.
-
Edit and copy
Edit the text in place, copy everything, or copy domains, URLs, emails and digit strings one by one.
Common use cases
Extract domains and links from screenshots
A colleague sends a screenshot full of domains: recognize it, then copy every domain, URL and email with one click — no retyping.
Quote chats and error messages
Turn conversation screenshots or error dialogs into editable text for docs, emails and tickets.
Recognize sensitive images safely
Contracts, IDs, internal dashboards — text is extracted without the image ever leaving your browser or passing through any third party.
Grab text on your phone
Open the site in a mobile browser, pick a photo from the gallery and copy the text. No app to install.
Tracking numbers and digit strings
Long digit strings like parcel numbers or order IDs are collected automatically, so you never cross-check by hand.
Local recognition vs. traditional online OCR
Most online OCR services upload your image to a server for recognition — often behind sign-up walls, usage limits and ads. PasteOCR moves the whole pipeline into your browser.
| What | PasteOCR | Upload-based OCR |
|---|---|---|
| Where recognition runs | Inside your browser | On a remote server |
| Where the image goes | Never leaves your device | Uploaded to their servers |
| Sign-up / login | Not needed | Usually required, or features gated |
| Usage and size limits | None | Often a few files per hour, 5–15MB each |
| Waiting in queue | None — instant once the model is cached | Queues at peak times |
| Ads and tracking | None | Common |
Frequently asked questions
Are my images uploaded?
No. Recognition is done by a WebAssembly engine inside your browser; images and results never leave your device. The server only serves the page and model files — there is no upload or storage endpoint at all.
Which languages can it recognize?
The engine is PaddleOCR's mixed Chinese–English model: Simplified Chinese, Traditional Chinese, English, Japanese and pinyin are recognized directly, with no language to choose. The interface itself is available in 10 languages.
Do I need to install anything or sign up?
No. Open the page and use it — no accounts, no usage caps, no watermarks, no paywall.
Why is a model downloaded on first use?
The recognition model is about 6MB in Standard mode. It is downloaded once, cached by your browser, and never fetched again on later visits.
How fast is recognition?
Typical screenshots finish within a few hundred milliseconds, depending on your device and image complexity. Images with a long side over 1600px are downscaled automatically to protect mobile memory.
Which image formats are supported?
PNG, JPG, WebP, BMP and GIF (first frame). Pasting, dragging and file picking all work.
Does it work on handwriting or stylized fonts?
Printed text and screen screenshots work best. Handwriting, artistic fonts and tiny low-resolution text are limited — try the Accurate mode.
Is anything kept after I close the page?
Results live only in the current page and disappear when you close it. The server keeps no images or text, and there are no analytics scripts.