Seedance's audio prompt feature lets you generate video that responds to audio — music, sound effects, or voice. The video's motion, timing, and visual elements sync with the audio, creating a more cohesive result than adding music in post-production.
This guide walks through the audio prompt workflow step by step.
Quick answer
Upload an audio file alongside your text or image prompt. Seedance analyses the audio for rhythm, beats, and energy, then generates video with motion that matches. Best results come from clear, rhythmic audio with distinct beats.
- You want video motion that syncs with music or sound effects.
- You are creating music videos, visualisers, or audio-reactive content.
- You want more dynamic video without manually editing to match audio.
How audio prompts work
When you provide audio alongside a prompt, Seedance analyses the audio waveform — identifying beats, tempo changes, volume dynamics, and energy shifts. It uses this information to drive the video's motion and visual intensity.
Loud, energetic sections produce faster motion. Quiet sections produce calmer visuals. Beat hits can trigger visual changes or motion accents.
Choosing the right audio
Clear, rhythmic audio works best. Songs with strong beats, obvious tempo, and dynamic range give Seedance the most information to work with.
Avoid audio with heavy compression (everything at the same volume), very complex layered music, or spoken word with long pauses.
- Electronic and pop music: strong beats, clear structure — works very well
- Orchestral music: dynamic range drives dramatic visuals
- Ambient music: produces subtle, slow-moving visuals
- Voice narration: motion follows speech patterns, pauses create stillness
Combining audio with text and image prompts
Audio prompts work best when combined with a text prompt that describes the visual style and a source image that establishes the scene. The audio drives the motion, the text guides the style, and the image anchors the visuals.
If you use only audio without a text prompt, Seedance generates abstract, visualiser-style content. Add a text prompt for more controlled output.
Settings for audio-driven video
Let the audio drive motion intensity rather than setting it manually. If you do set motion intensity, use medium values (40-60%) — too low and the audio sync is not visible, too high and the video becomes chaotic.
Match the duration to the audio length supported by the current platform. Start with a short section of your audio that captures the energy you want, then extend only after the timing survives review.
Editing audio-driven clips
Generate multiple clips from different sections of the same audio track and edit them together for a longer piece. This works well for music videos — each clip captures a different section's energy.
In post-production, you can extend or loop clips, add transitions, and overlay text, but treat the generated audio-visual sync from Seedance as a draft foundation that still needs review.
Worked example: music visualiser clip
You have a 4-second clip from an electronic track with a strong beat drop. Upload the audio with a text prompt: 'Abstract neon landscape, camera pushes forward, energy surges with the music, vivid colours, cinematic.' Seedance should generate a clip where the visual motion and intensity follow the beat drop closely enough to review.
Common mistakes
- Using audio with no dynamic range — the video will have no motion variation.
- Not adding a text prompt, resulting in generic abstract visuals.
- Trying to generate long clips — sync and quality can drift depending on the current mode.
Step by step: write audio-aware prompts
- Separate the audio line from the visual line. "Visual: a chef slicing vegetables. Audio: rhythmic knife chops, kitchen ambience." Seedance handles split prompts better than blended ones.
- Name one ambient layer. "Coffee shop hum" or "rain on window". One layer is clean; three layers fight each other.
- Name one foreground sound. Footsteps, a bell, a laugh — the sound the viewer should notice.
- Treat dialogue as a review item. ByteDance describes Seedance as supporting audio-video generation, but clean speech and word-level sync still need testing in your current platform. For client-ready dialogue, generate or record the voice separately unless your short test proves the built-in output is good enough.
- Match sound timing to motion. If the clip is 5 seconds, describe a 5-second audio event ("a single door slam at the 3-second mark").
- Export at 48kHz for editing. Keeps sync accurate when you layer in extra audio later.
Troubleshooting table
| Symptom | Likely cause | Fix |
|---|---|---|
| Audio feels disconnected from visual | Prompt described visual only | Add an explicit audio line. |
| Ambient sound drowns out the action | Ambient layer too dominant in prompt | Describe ambient as "quiet" or "low". Foreground sound should be the main descriptor. |
| Dialogue sounds garbled | Speech generation or lip-sync quality is not strong enough for this clip | Record voiceover separately, or rerun a short audio-enabled test before using built-in speech. |
| Sync drifts by the end of the clip | Clip is too long or the selected mode does not support that timing well | Use a shorter supported duration and check the live platform settings. |
For the beginner workflow, start with the Seedance tutorial. For lip-sync specifically, see lip-sync and talking heads.
When to use something else
For standard video generation without audio, see Seedance image to video. For YouTube content creation, see Seedance for YouTube Shorts.
Frequently asked questions
How does Seedance use an audio prompt?
It analyses the waveform — beats, tempo, volume and energy shifts — and drives the video's motion and intensity from that. Clear, rhythmic audio with a strong beat gives it the most to work with; flat audio with no dynamic range produces little motion variation.
Should I set motion intensity, or let the audio drive it?
Let the audio drive it where you can. If you do set it manually, use medium values around 40-60 — too low and the sync is not visible, too high and the clip becomes chaotic.
Do I still need a text prompt if I am using audio?
Yes. Audio alone produces abstract, visualiser-style output. Pair it with a text prompt for the style and a source image for the scene: the audio drives motion, the text guides style, the image anchors the visuals.
What is the best way to write an audio-aware prompt?
Split it into a visual line and an audio line — 'Visual: a chef slicing vegetables. Audio: rhythmic knife chops, kitchen ambience' — and keep it to one ambient layer plus one foreground sound. Blended prompts and three competing layers both confuse the model.
Can I rely on Seedance for clean dialogue and lip-sync?
Treat it as a review item, not a guarantee. ByteDance describes Seedance as supporting audio-video generation, but word-level speech sync still needs testing on your current platform — for client-ready dialogue, record or generate the voice separately until a short test proves the built-in output is good enough.
How do I keep audio and video in sync across a longer piece?
Generate short clips from different sections of the track and edit them together rather than one long clip — sync tends to drift on longer renders. Match each sound event to the motion timing, and export at 48kHz so the sync holds when you layer audio in post.
Related guides on this site
These guides cover image-to-video, content creation, and prompt writing for Seedance.