Mood Video
summary
A concrete system that generates short mood videos from webcam and microphone input shows a browser-only pipeline and a live demo that actually runs. It uses TensorFlow.js emotion model for face expressions, an audio sentiment extractor and ffmpeg.wasm for editing, assembles 15–30s clips, and exposes an npm package. Evaluation on 120 generated clips from 30 participants reports 78% accuracy for mood labels and sustained real-time throughput around 10–15 FPS so the prototype works end to end.
AI-written summary · about
