Applies to: Only video models that generate audio with the video. Not all video models produce sound.
Make sure the video model allows it. Enable "Generate Audio" or "Audio" toggled On in the video settings input window.
Use these four techniques to control the sound, dialogue, and music in a generated video.
Name specific sounds in the video prompt: "footsteps on gravel," "rain on metal," "crowd murmur" — or state "silent" deliberately.
Put spoken lines in double quotes to signal lip-sync. Keep lines short — as a guideline, one short sentence per 5 seconds of video.
Describe the voice: "soft female voiceover," "gravelly whisper."
For music, describe genre, instruments, mood, and timing: "soft and mysterious for the first 10 seconds, crescendo in the final 5 seconds."
On Seedance 2.5 you can also score audio with a dedicated syntax: ( music ) < effects/ambience > { dialogue } [ subtitles ], assigning the speaker, language/accent, and delivery. See Seedance 2.5: advanced reference & editing controls.