Why your clips look ready-to-ship
Engineering choices that separate StreamSnip from a generic clip tool.
Word-accurate caption timing
Most tools use sentence-level timing — captions appear all at once and stay for a whole line. We use Whisper for per-word timings and burn captions with cinematic word emphasis (bold, color, size). That's the TikTok-native look that outperforms static captions.
Face-tracked reframing
For 9:16 vertical exports, we track the streamer's face throughout the clip and keep them in the safe zone. When we detect a facecam over gameplay, we auto-generate a split-screen layout (face on top, gameplay below).
Clean audio encode
Every clip re-encodes audio with dialog-forward levels. Full LUFS / EBU R128 mastering is on the roadmap; until then, output is broadcast-friendly without being guaranteed broadcast-spec.
Multi-signal scoring
Virality score is a weighted blend of audio peaks (40%), chat velocity (30%), clip duration fit (15%), and emote density (15%). Non-Twitch sources fall back gracefully when chat isn't available.