← Back to VideoOCR
Experiment
Cloud model test

Upload a game-chat recording and send it straight to a vision model — no local OCR. Output is deduplicated verbatim chat text for comparing against the PaddleOCR pipeline.

qwen3.8-flash
🎬

Drop a video here or browse

MP4 · MKV · MOV · AVI · WEBM · 500 MB MAX

The video is downscaled to 480p @ 1 fps before sending. Videos over 10 minutes are trimmed to the first 10 minutes for this test. Usage tokens are reported with the result.

Preparing…
Something went wrong