Manga page analyzer (Magi v3 + text detector + manga-ocr)
Panels, characters, texts, tails + OCR -> JSON. Magi licence: non-commercial.
Manga page
Drop Image Here
- or -
Click to Upload
Detections
Text language (ja = detector + manga-ocr, en = Magi OCR)
ja
en
Convert page to grayscale first
Probe Magi's own OCR (predict_ocr) and show raw output
Use proximity to the nearest character when no tail is found (marked medium/low confidence)
Analyze
Result
API: base64 endpoint (used by the Android app)
▼
Base64 image (jpg/png)
lang
ja
en
proximity speaker fallback
Result
Run