H3 lipsync tip: adding lyrics to the prompt drastically improves mouth movements
redkinoko · reddit · 2026-10-20
Reddit user redkinoko shares a practical finding about H3 lipsync: driven by audio alone, the model guesses mouth shapes without knowing the actual words, causing mismatched mouth movements on some sounds. Including the sung lyrics in the prompt improves the motion drastically.
Their automated workflow splits a song into 10-second segments, so the prompt for each segment must be modified to match exactly the lyrics being sung. They suggest a possible automation step — feeding segmented audio to ChatGPT for lyric extraction, then injecting it into the ComfyUI workflow prompt — though this remains untested.
More from coding & agent
- Stanford researchers study cascade harnesses to maximize agent performance at low cost — james_y_zou · 2026-10-20
- Dev ran a Shopify store end-to-end with Hermes-Agent, making physical merch — alexcovo_eth · 2026-10-20
- LlamaIndex launches ExtractBench Pro: long-tail document extraction that stumps frontier models — llama_index · 2026-10-20
- vLLM ecosystem ships d3-turbo: multimodal decision model, 15x faster, tops Jev Decision Index — vllm_project · 2026-10-20
- Compacting a compacted LLM context: engineers' silent pain — nrehiew_ · 2026-10-20
- Harness over new models: Turing Post video on closing the capabilities overhang hits 60k views — TheTuringPost · 2026-10-20