Stability AI
Stable Audio 2.5
Instrumental music and sound up to three minutes long. On Leap, it sits beside 130 other models in one studio.
- Music
- Up to 190 s
- Start free, no card
- Every top model, one account
- Pay per run, priced before it starts
- 01Write the words, pick a voice
- 02See the exact price, then press go
- 03Keep it in your library, or take it further
Start with an idea
Three ways to try Stable Audio 2.5
Questions
Stable Audio 2.5 on Leap, answered.
- Is this the real Stable Audio 2.5?
- Yes. Leap sends your request to Stability AI's model through its provider and gives you back what it makes. Switch to any other model in the same studio when an idea needs it.
- What does it cost?
- In dollars, per run: the provider's list price plus 10%. You see the exact price before every run, a run that fails costs nothing, and there is no subscription. Every model's price is on the pricing page.
- Do I need a Stability AI account?
- No. One Leap account and one balance run Stable Audio 2.5 and every other model, in the studio, through the API and from your agents.
- What happens to what I make?
- It stays in your library while your account is active, ready to use anywhere.
Your idea, Stable Audio 2.5.
Start free, no card. Every run shows its price before it starts.
Make it with Stable Audio 2.5For developers
Run Stable Audio 2.5 from code
- Model ID
- stability-ai/stable-audio-2.5
- Makes
- Voice and sound
- Pricing
- Per run, quoted first: POST /v1/quotes
Inputs
What the model takes. Leave out anything with a default.
Promptprompt
Text, up to 4,000 characters, required
Durationduration
1 to 190, required. Default: 190 s
Num Inference Stepssettings.num_inference_steps
4 to 8. Default: 8The number of steps to denoise the audio for
Guidance Scalesettings.guidance_scale
1 to 25. Default: 1How strictly the diffusion process adheres to the prompt text (higher values make your audio closer to your prompt).
Seedseed
0 to 4294967295The same seed and input give the same image.
The request
The same envelope for every model; only the model and its input change. With prefer: wait=60 the answer holds the result. Agents can call it through the MCP server at mcp.tryleap.ai.
curl https://api.tryleap.ai/v1/generations \ -H "x-api-key: $LEAP_API_KEY" \ -H "content-type: application/json" \ -H "prefer: wait=60" \ -d '{ "model": "stability-ai/stable-audio-2.5", "input": { "prompt": "A paper boat drifting down a rainy street at dusk" } }'