Image to text (OCR)

Choose once; switch tools freely.

Choose an image and recognize text without selecting a language. Images stay in your browser.

How to recognize text

Choose one static JPEG, PNG or WebP. Keep photographed text clear, flat and free of glare; orient large rotations before importing.
Recognition starts automatically after image validation. Enhanced recognition falls back to compatible English, Chinese and Japanese recognition if unavailable or unsuccessful.
Proofread the actual text. Select/copy it or download UTF-8 TXT.

Proofread the text, numbers and punctuation before use.

Images and text stay in your browser. No upload or saved history.

JPEG / PNG / WebP · up to 15MB, 4MP, 4096px per side

Common questions

Does recognition upload my image?

No. After you select an image, this site supplies the fixed models needed for automatic recognition. Images and text are processed in a task-owned browser Worker, with no cloud OCR or local server. Enhanced models total about 31MB; compatible models about 24MB.

Which languages does enhanced recognition support?

The PP-OCRv6 small model is published for 50 languages, including Chinese, Japanese, English and Latin-script languages such as French, German and Spanish. This is its training scope, not a guarantee for every language or image. Compatible mode supports English, Simplified Chinese, Traditional Chinese and Japanese. Proofread names, numbers and punctuation; blur, glare, perspective, handwriting and complex layouts can still cause errors.

What are the processing limits?

JPEG, PNG or WebP up to 15MB, 4MP and 4096px per side. Each task has a 45-second deadline including downloads. Enhanced processing caps allocated GPU texture payloads at 128MiB; compatible processing caps its linear heap at 256MiB. These are not total physical-memory guarantees. Output is at most 100000 UTF-16 units / 400000 UTF-8 bytes. Cancel, edits and leaving terminate the Worker and clear old results.