Extract Text from Image

Makers comment: Hi Fazier! You know that moment when the text is right there on the screen but you cannot select it? That is what this fixes. Paste or upload an image and get copyable text in a second. The part I cared about most: recognition runs in your browser, so screenshots of private chats or documents never go to a server. It is free with no account. Tell me which language it reads worst for you, that is what I am tuning next.

Running the OCR on the device is the reason I'd pick this over the online ones. Two things I'd use a lot: right-click "copy text from image" on any image on a page, so I don't have to screenshot first, and keeping line breaks for screenshots of code, since error messages and stack traces lose their shape when they come out as one paragraph.

Hey Fazier community! Ever had to retype text from a locked PDF, video frame, or image just because you couldn’t highlight it? I built this to eliminate that hassle. Just drop in your screenshot and grab the text instantly. What mattered most to me while building this was privacy: the OCR processes 100% locally on your machine—nothing gets sent to the cloud, so your sensitive notes, receipts, and internal chats stay completely yours. No sign-ups, totally free to use. Test it out and let me know if it stumbles on any specific fonts or formats so I can polish the next update!

Running the recognition on the device is the right call for receipts and invoices, which people don’t want to upload anywhere. How does it handle a column of amounts, say a photo of a receipt or a table: do the numbers come out in their rows with the decimal points intact, or as one block of text?

Running the recognition on the device is the right call for receipts and invoices, which people don’t want to upload anywhere.

good work and good work

Running OCR on-device so images never leave the computer is a strong selling point, especially for receipts, invoices and private documents where people hesitate to use upload-based tools. Which OCR engine is behind it, and how does accuracy hold up on low-quality phone photos compared to clean screenshots?

Running the OCR on the device is the right call for receipts and private documents. How does it handle tables, where the columns should stay lined up in the TXT? A toggle between keeping line breaks and joining lines into paragraphs would save a cleanup step when copying from slides.

Super useful tool! Love that it supports multiple languages and keeps everything private by processing images directly on your device.

The paste-with-Ctrl+V path is the one I would actually use: screenshot an error, paste, then download as .txt. I am curious whether a two-column scanned page keeps reading order, or whether lines from the right column get mixed into the left.

Useful product Extract text from image

Running the OCR on-device is the detail that makes this stand out — screenshots of chats, error messages and receipts often contain private info I would not want to upload anywhere. The Ctrl+V paste flow also fits how people actually work: screenshot, paste, copy. I am curious how it handles low-contrast screenshots or handwritten notes, since that is where most OCR tools struggle. Congrats on the launch!

The paste-with-Ctrl+V path is the one I would actually use: screenshot an error, paste, then download as .txt. I am curious whether a two-column scanned page keeps reading order, or whether lines from the right column get mixed into the left.

Super useful tool! Love that it supports multiple languages and keeps everything private by processing images directly on your device.

On-device OCR is the right default for a screenshot of an error dialog or a chat, because those images often include account details you should not upload. Paste with Ctrl+V is also the path I would use, since the image is already on the clipboard from the snip tool. The harder case is a photo of a receipt, where the numbers are small and the paper is tilted. Does the extension still keep that photo local when you upload a file, and is there a way to correct a misread digit before the TXT download?

On-device OCR is the right call for this use case - screenshots often contain exactly the kind of content people should not upload to a random server. Curious how it handles dense CJK text: Japanese and Korean are listed, how about Chinese? For a next step, a bulk mode that ingests several screenshots at once would fit the receipt-and-invoice use case nicely.

Does it handle vertical Japanese text? Manga panels and a lot of signs here are written top to bottom, and many OCR tools read them as horizontal lines and scramble the order.

This type of product is really good for our company!

On-device OCR is the right call for screenshots of chats and invoices. How does it handle multi-column layouts, like a two-column page or a slide with several text boxes? Does it keep the reading order, or mix lines across columns? A "copy as table" option for receipts would also be really useful.

Doing the OCR in the browser is a solid choice. Out of curiosity, what's the engine underneath: Tesseract compiled to WASM, or something ONNX-based? And how heavy is the first language download? That startup time usually decides whether people keep an extension like this installed.

The on-device approach is what makes this practical for everyday work. Being able to paste a screenshot with Ctrl+V and copy the text, or save it as a .txt file, would save a lot of time when pulling an error message or a figure from a slide into notes. It would be useful to know how it handles low-contrast screenshots, since those are often the ones that are hardest to retype accurately.

Running OCR entirely on-device is a strong privacy angle, particularly for screenshots of receipts and private chats. Have you considered a mode that preserves table or column layout, so copied numbers from invoices stay aligned in the TXT export?

Running the OCR on the device is the reason I'd pick this over the online ones. Two things I'd use a lot: right-click "copy text from image" on any image on a page, so I don't have to screenshot first, and keeping line breaks for screenshots of code, since error messages and stack traces lose their shape when they come out as one paragraph

On-device OCR is the right default for a screenshot of a private chat or a receipt you do not want uploaded. When I copy numbers off a photo of an invoice, the hard part is usually keeping line breaks and columns instead of one flat paragraph. Does the TXT download preserve those line breaks, or does it collapse the page into a single string?

Running OCR on-device is the part that would make me trust this for receipts and screenshots of private chats. One question: when you copy from a screenshot that has columns or a receipt layout, does it preserve line breaks and rough layout, or does it flatten everything into plain running text?

good work keep it up

Research and analyze the project requirements, structure, and technology, then implement the necessary changes.

For screenshots containing two languages, can recognition detect both automatically, or should the user choose language packs first? Showing that choice before extraction would help with multilingual slides and documents.