Report
Tips for keeping audio references the same in SeeDance 2/2.5
A user recommends matching reference and generation lengths, providing a transcript and telling SeeDance not to change the audio.
TLDR
A user shares tips for keeping audio references the same in SeeDance 2/2.5: match the reference length to the generation length, provide a transcript, describe the audio in the prompt (for example, “lipsync”) and tell the tool not to change the audio. They suggest generating in smaller chunks, saying 10 or 15 seconds of dialogue is more likely to stay the same than 30.
Combined views
—
2 Sources, first seen ago
