Add sound effects and music to a silent video
Footsteps, rain, doors and a score, in time with the picture
A clip from an AI video model, a screen recording or a drone shot often arrives silent, and silence reads as unfinished. Seed Audio 1.5 takes the video and a short description of what should be heard, and returns a soundtrack that follows the action on screen.
Based on ByteDance's documentation and prompt guide. Our own published test is the translation demo on the Seed Audio 1.5 page.
Describe it, or place it
ByteDance's prompt guide offers three levels of control. Most clips need only the first.
A plain description
“Heavy rain on a tin roof, distant thunder, no music.” Good for ambience and simple scenes.
An audio script
The order of events: the door opens, footsteps cross the room, a chair scrapes, music swells.
A timeline
Exact seconds for each sound, [0.4s:1.0s] for a click, [12.0s:15.0s] for a music lift, so effects land on your cuts.
Typical jobs
AI-generated clips
Give generated footage the footsteps, wind and room tone it is missing before it goes into an edit.
Product and UI videos
Clicks, swipes and a light music bed placed on the exact frames of the interaction.
Game and trailer scenes
Magic hums, impacts and rising drums, built from a description instead of a sound library search.
Social shorts
A music bed with a hit on the cut, without licensing a track.
Limits
| Parameter | Limit |
|---|---|
| Video length | 4 to 360 seconds |
| Formats | MP4, MOV |
| Resolution | 480p, 720p or 1080p |
| File size | Up to 200 MB |
| Music length | Up to 6 minutes |
| Output | WAV (default), MP3, PCM, OGG Opus |
Source: ByteDance Seed Audio 1.5 early access integration guide, October 2026.
Prompts to start from
Copy one, swap in your own scene. The examples are in English, which is the language the model follows best for stage directions.
A single effect
A heavy metal pipe falls onto a metal surface: a dull, low, weighty impact with a dense metallic thud. Close and clear against a quiet background. No speech or music.
Describe the object, the surface, the size of the sound and what must be absent.
Ambience for a street scene
Realistic old-town street at dawn: distant church bell, a few footsteps on cobblestones, a cart wheel creaking past from left to right, light wind. No speech, no music.
Directions like “left to right” help the sound follow movement on screen.
Timed music and effects
[0.0s:20.0s] Calm ambient synth pad under the whole clip. [3.2s:3.6s] Soft whoosh as the camera turns. [9.0s:9.4s] A single bright chime. [16.0s:20.0s] The pad fades out.
Use timestamps only where a sound must sit on a specific frame.
Keep in mind
- Describe what you want to hear. The more specific the object and the surface, the more specific the sound.
- ByteDance's prompt guide recommends stating what must be absent, such as “no speech” or “no music”.
- Long films go scene by scene: one request covers up to 6 minutes.
FAQ
Can AI add sound effects to a video automatically?
Seed Audio 1.5 generates effects, ambience and music that follow the video, from a description you write. It needs that description: it does not decide on its own what the scene should sound like.
Can I control exactly when a sound plays?
Yes. Put timestamps in the prompt, for example [3.2s:3.6s], and the sound is placed in that window.
Can it add music too?
Yes, as part of the same soundtrack, or as a separate music generation of up to 6 minutes.
What videos work?
MP4 or MOV, 4 to 360 seconds long, 480p to 1080p, up to 200 MB.
Related
Make the silence sound finished
Describe the scene, pin the key moments, get one soundtrack.
Add sound to a video