Remove image background locally
Remove backgrounds locally with a fast portrait model or a finer general model, then reuse the mask with transparency, a solid colour, or another image.
Image to text (OCR)
Recognize text · Select a region · Export text
Images stay on your device
Files
0 · 0 selected
No images selected yet.
Drop images or a ZIP file here
You can also paste an image or select several at once.
JPEG · PNG · WEBP · AVIF
Settings
Advanced settings
Export settings
ID photo compositionOn-device detection only suggests a crop; you can still adjust it before export.
This image’s crop has not been confirmed.
ID photo backgroundOptional on-device portrait cut-out; the original never leaves your device.
Dimensions
Preset sizeCommon platform sizes
Select one or more sizes; each image is exported in every selected size.
Icon safe area
One composition generates favicon, PWA and common app sizes, plus a manifest.
Crop areaDrag the box to resize or reposition it
Rotation0°
Quality
Padding on each sidepx
Upscale imageReal-ESRGAN · 2×
Pick a colour from the image
Frame rangeLeave the last frame blank to use the rest.
Base64 increases data size. This tool limits input to protect browser memory.
About this tool
Extract text from screenshots, photographed documents or scans by choosing the language in the image. Crop to the relevant text when helpful, then verify amounts, dates and identifiers before copying or exporting.
How to use
Image to text (OCR)
What this tool is for
Recognise Simplified Chinese, Traditional Chinese, or English text locally and export readable text plus structured JSON with confidence data.
How to use it
Select the language actually present in the image and crop to the text region when useful, then run local OCR. The result includes copyable text, confidence, and structured data; tables, handwriting, curved text, and low-resolution scans require proofreading.
Results and next steps
Results include copyable plain text, confidence data, and a structured representation. Proofread each section before formal use.
Privacy and local processing
Processing runs in your browser. Source images are not uploaded by this tool.
Common questions
Does OCR support every language available in the page interface?
No. Current OCR models are English, Simplified Chinese and Traditional Chinese; choose the language in the image. TXT contains plain text; JSON also includes confidence scores and TSV data. The tool does not automatically reconstruct Excel spreadsheets.
Can I skip proofreading when confidence is high?
Confidence is the model's estimate, not a measured character-by-character accuracy score. Check amounts, dates, identifiers and easily confused characters. Low-resolution images, handwriting, curved text and complex tables need particular care.
How can I improve missing text or incorrect reading order?
Choose the actual text language and crop to one column or the required passage. Correct tilted document photos in the scanner first. Use a clear original rather than repeatedly compressed screenshots; columns, tables and handwriting still need manual review.
Related tools
Opens in a new tab
Comments
0