By generating a clean line with Gemini 3.1 Flash TTS and passing that WAV to Seedance as reference_audios, you get noticeably better Seedance audio. On longer clips with a lot of narration, Seedance's bare-prompt audio tends to garble — words mush together, sentences slop into each other, the take is unusable. Handing it a reference WAV of the exact line fixes that. This workflow demonstrates the technique simply.
- 1 clip
- 1.2 MB
Generate short narrated explainer videos in a minimal hand-drawn stick-figure style — black outlines on grey textured paper, AI-narrated, about 40–60 seconds long. Good for educational content where a concrete visual metaphor carries the explanation: how things work, why mechanisms behave the way they do, what's actually happening inside a process you usually just hear about.
- 8 clips
- 10.4 MB