SnapBit Tools

Image to Text

Upload image

Drop a photo, scan, or screenshot here, choose a file, or paste an image with Ctrl+V.

Runs on your device with the Mirage model (20.6 MB, downloaded once). Your image is never uploaded.

Mirage reads Latin-script text (English, German, French, Spanish, Italian, Dutch, Polish and similar). Small or stylised text can be misread. Check important results.

Free Image to Text (OCR) - Extract Text from Images Privately

Extract text from photos, scans, and screenshots for free with SnapBit's Image to Text tool. Drop or paste an image and get editable text in reading order, with every line boxed on the image and scored for confidence. Nothing is uploaded: the OCR runs on your device. Most online OCR services send your image to a remote server. That is a poor fit for receipts, contracts, ID documents, and internal screenshots. This tool runs Mirage, an open-source 51M-parameter vision model from Cactus Compute, on the Cactus WebAssembly engine inside a Web Worker in your browser. The model (about 21 MB) downloads once from Hugging Face and is cached, so later runs work offline. Mirage does more than read text. Switch to Describe & detect to get a scene summary, a list of objects with bounding boxes, segmentation masks and attributes such as colour and shape, and the relations between them, in seven languages. Use Ask questions to fill a small form about the image: choose between labels you supply, count things within a range, or read a short field, each with its probabilities. Mirage reads Latin-script text, which covers English, German, French, Spanish, Italian, Dutch, Polish, and many other languages. Clear printed text gives the best results. Small, blurry, or stylised text can be misread, so you can edit the result before copying it.

Key Benefits

  • ✓OCR that runs entirely in your browser, with no uploads
  • ✓Paste screenshots straight from the clipboard
  • ✓Editable text in reading order, ready to copy or save as TXT
  • ✓Line-by-line confidence scores and boxes on the image
  • ✓Scene description, object detection, and segmentation masks
  • ✓Ask questions about an image and get answers with probabilities
  • ✓Works offline after the one-time 21 MB model download

Common Use Cases

  • ✓Copying text from screenshots, slides, and error dialogs
  • ✓Digitising receipts and printed notes without a cloud service
  • ✓Reading text from photos of signs, labels, and packaging
  • ✓Pulling quotes from scanned pages and book photos
  • ✓Describing images and generating alt text drafts
  • ✓Sorting images by type, such as receipt, letter, or photo
  • ✓Privacy-sensitive OCR for documents that must stay on your device

Frequently Asked Questions

It runs Mirage, an open-source vision model from Cactus Compute, on the Cactus WebAssembly engine in a Web Worker in your browser. Mirage finds text regions, reads each line, and returns it with a bounding box and a confidence score.

No. Your image is decoded and analyzed on your device. The only network request is the one-time download of the Mirage model (about 21 MB) from Hugging Face, which your browser then caches.

Mirage reads Latin-script text: 306 characters covering English, German, French, Spanish, Italian, Dutch, Polish, and many other languages that use the Latin alphabet. It does not read Chinese, Japanese, Arabic, Cyrillic, or other scripts.

Clear printed text, such as documents, receipts, and screenshots, reads well. Small, blurry, curved, or handwritten text can be misread. Each line shows a confidence score, and you can edit the text before copying it.

Describe & detect returns a scene summary, objects with boxes, segmentation masks and attributes, and relations between them, in seven languages. Ask questions fills a small form: it picks from labels you give it, counts within a range, or reads a short text field, with probabilities for each answer.

The first run downloads the Mirage model and starts the WebAssembly engine. Later runs use the cached model and usually take a few seconds per image, depending on your device.

100% private | Nothing leaves your device

Your data never leaves your browser. All processing is done locally on your machine, ensuring maximum security and speed. No server calls, no tracking, no risk.

Privacy Guaranteed

Other Image tools

View Image tools