11 — AI Audio & Music Generation
AI Audio Generation lets you create original background music, instrumentals, and sound effects using AI models. Generated audio can be downloaded or added directly to your timeline as an audio clip.
Access this feature under AI Generate → Audio.
A Replicate API key is required for all models.
Available Models
Stable Audio 2.5
Best for: instrumentals, sound effects, loops
| Setting | Description |
|---|---|
| Prompt | Describe the sound or music style |
| Duration | 1–190 seconds |
| Steps | Generation quality (4–8) |
| CFG Scale | Prompt adherence (1–25) |
| Seed | Optional, for reproducibility |
Example prompt: "Upbeat corporate background music, piano and strings, 120 BPM"
Google Lyria 2
Best for: high-quality 48kHz stereo music
| Setting | Description |
|---|---|
| Prompt | Music description |
| Negative Prompt | What to exclude (e.g., “vocals, drums”) |
| Seed | Optional |
MiniMax Music 1.5
Best for: full vocal songs with lyrics
| Setting | Description |
|---|---|
| Prompt | Genre, mood, and style description |
| Lyrics | Full song lyrics with structural tags (see below) |
Lyrics tag format:
[verse]
Your verse lyrics here
[chorus]
Your chorus lyrics here
[bridge]
Optional bridge lyrics
ElevenLabs Music
Best for: studio-quality music production
| Setting | Description |
|---|---|
| Prompt | Music description |
| Duration | 5,000–300,000 milliseconds (5 seconds to 5 minutes) |
| Force Instrumental | Toggle to prevent any vocal generation |
ACE-Step
Best for: open-source full songs with custom lyrics
| Setting | Description |
|---|---|
| Tags | Style descriptors (comma-separated genre/mood tags instead of a prompt sentence) |
| Lyrics | Full lyrics text |
| Duration | Clip length in seconds |
| Guidance Scale | Prompt adherence |
| Number of Steps | Generation quality |
| Seed | Optional |
Example tags: "pop, upbeat, female vocals, guitar, 120bpm"
Meta MusicGen
Best for: melody-conditioned generation — provide a reference melody
| Setting | Description |
|---|---|
| Prompt | Description of the desired music |
| Duration | Clip length |
| Temperature | Creativity (higher = more variation) |
| Top K | Sampling diversity |
| Input Audio | Optional reference audio to condition the melody |
| Guidance Scale | Prompt adherence |
| Seed | Optional |
Generating Audio
- Select your model.
- Fill in the prompt and any model-specific fields.
- Click Generate.
- The job runs in the background via the Job Queue.
- When complete, the result appears in the Results tab with an audio player.
Using Generated Audio
From the Results tab:
– Click Play to preview the audio in the browser
– Click Download to save the MP3 file
– Click Add to Timeline to insert the clip into the audio track at the current playhead position
Previous: Talking Head / Lip-Sync | Next: Long-Form Video Pipeline →