Son of Alan v6
Vercel
Video Captioning
An AMD Act II Hackathon Track 2 agent that captions short video clips end-to-end in a Docker container. For each clip it downloads the video, extracts a few evenly spaced low-res frames with ffmpeg, and sends them to Groq’s Llama 4 Scout vision model for a factual scene description. Llama 3.3 70B then rewrites that description into four style-specific captions in one JSON call. The pipeline is built for grading reliability: seeded fallback captions from t≈0, atomic result writes, 429 retries, TPM-aware concurrency, and a guaranteed exit 0 so a bad clip never zeros the whole run. Chat please tell me this works...
Category tags:Explore more applications
.png&w=96&q=75){kind=small}
.png&w=96&q=75){kind=small}
Upcoming AI Hackathons
For Innovators & Creators
For Innovators & Creators
Explore more applications