VLM OCR + FIELD EXTRACTION ENGINE
Document review station
Compare the camera image with processed output. Select a line or field to inspect its evidence box, line number and JSON trace.
SYNTHETIC FIXTURESTELEOCR + PP-OCRv6
DEMO
Use synthetic records for live uploads. Live files pass through this Worker to a protected Modal GPU endpoint. This demo stores no upload, but its processing path is cloud-hosted and does not meet the scope's self-hosted, no-egress requirement.
LIVE MODAL INFERENCE
Process up to 8 pages
Choose page images or a PDF. PDFs are rendered in your browser and the first 8 pages are sent as images. The current endpoint has no long-file queue or resume support.
-test pages
-document classes
-largest template schema
-PP-OCRv6 sample CER
2languages
1conflict fixture
Sample set drawn from the latest Scope of Work
SYNTHETIC TEST PAGE
Loading examples
01
ORIGINALCamera input
-PHONE CAPTURE
02
Processed output
READING ORDER0 LINES