Multilingual OCR: Choose the Recognition Language

To extract Chinese, Japanese, Korean or another supported language, open Image to Text and select the language used in the image. English is selected initially, regardless of the website language. There is no automatic-language detection or combined-language option.

Available languages

The 15 choices are English, Simplified Chinese, Traditional Chinese, Japanese, Korean, French, German, Spanish, Italian, Portuguese, Russian, Arabic, Hindi, Thai and Vietnamese. A language absent from this list is not an option on the current page.

Recognize one image

1. Select a JPG/JPEG, PNG, WebP, GIF, BMP or TIFF. Selecting the image immediately starts uploading it to the server.
2. Choose the recognition language, including the correct Chinese writing system when relevant.
3. After the upload finishes, click Extract Text to start recognition on the server.
4. Compare the result with the source. Copy it or download the UTF-8 TXT file; paste into an editor to make corrections.

For PDF content, first use PDF to Image, then recognize the required page images individually.

Input limits

Each task accepts one image up to 20 MB. Each side must be at most 8192 px, and width × height must be at most 25 million pixels. Only the first frame or page of a GIF or multipage TIFF is recognized. These limits apply together.

Tips for Asian Language OCR

Chinese text

Choose Simplified or Traditional Chinese to match the source. Check similar-looking characters, punctuation and names against the image. Keep full strokes visible rather than shrinking the image until dense characters merge.

Japanese text

Choose Japanese for text containing Kanji, Hiragana and Katakana. Small furigana can be confused with the main line; inspect its position and reading order. Vertical columns and mixed English labels require proofreading and may need separate crops.

Korean text

Choose Korean and check the spacing between Hangul syllable blocks. Use a clear source so individual strokes do not merge. English brand names or numbers in the same image still need checking; there is no guaranteed accuracy for mixed scripts.

Tips for the Best Results

Image Quality

Clear, high-resolution images work best for all languages.

Font Clarity

Standard fonts are recognized better than decorative ones.

Background Contrast

High contrast between text and background improves accuracy.

Text Size

Make sure characters are large enough to be clearly visible.

Common Use Cases

Translation Preparation

Extract foreign-language text for translation services.

Document Digitization

Convert printed foreign-language documents into digital text.

Travel Photos

Extract text from signs, menus, and documents while traveling.

Research

Capture text from foreign-language academic sources.

Business Documents

Process multilingual business correspondence and contracts.

Mixed languages and difficult layouts

For a mixed-language page, choose the main language and check every section. Recognition of other scripts is not guaranteed. You can crop separate text regions with Image Cropper and recognize them individually with different language choices.

Keep characters sharp and upright. Handwriting, vertical text, decorative fonts, small annotations and columns may need correction. The tool produces plain text, not a translated document, formatted table or searchable PDF.

Check accuracy yourself

There is no fixed accuracy percentage for a language. Image quality, lettering and layout affect each result. The displayed model confidence is not a measured percentage of correct characters. Check names, dates, amounts, punctuation and reading order against the source.

Troubleshooting

If output uses the wrong script, check the selected language first: the tool does not detect it automatically. If characters are missing, compare source sharpness, contrast and crop boundaries. Separate dense columns or small annotations where useful. Try the same short sample after one change so you can judge whether it helped; repeatedly submitting an unreadable original is unlikely to resolve the underlying problem.

FAQ

Does selecting Japanese translate the text?

No. It chooses a recognition model; the output remains in the source language.

What if the language is missing or the result is poor?

Use a workflow that supports the required language or layout. For supported languages, try a clearer source or a tighter crop and proofread again.