id-doc-reader
Info
📷
Choose a capture method, then tap
Start
.
Capture method
Auto — smart reader (detector + countdown)
Manual — shutter button only
Scenario
MrzOrLocate — MRZ fast-path + locate
Locate only (no MRZ model)
MrzAndLocate — strict, MRZ required
Use MRZ detector (uncheck to fall back to row-darkness heuristic)
Start
Stop & reconfigure
Capture result
Cropped (ID-1 warp)
Raw frame
Backend response
Scan again
Uploading to backend…
Sending the full camera frame and waiting for the response. Don't close the tab.
0.0 s
Backend URL (blank = skip upload, client-only test)
Debug mode (live mask preview)
Hide RAW preview after capture (saves ~4 MB DOM bitmap on iOS)
model input (320x200 RGB)
raw mask (sigmoid)
MRZ debug — strip with extent + actual model inputs
strip (red lines = detected MRZ extent)
Last capture
(no capture yet)
CROPPED (ID-1 warp)
RAW (full frame)
Response
(none)
Events