Short answer: Stable Audio 3 Medium makes instrumental music and sound effects up to 380 seconds for $0.035 a request at any length; Stable Audio 2.5 makes the same kinds of clips up to 190 seconds for $0.68 a request; ACE-Step 1.5 XL is the one for songs with sung lyrics, from 10 to 240 seconds at $0.00025 a second ($0.0075 for 30 seconds). Listen to the clips below and pick by the sound you need. All three run on EmpirioLabs through one API.
Specs at a glance
| Stable Audio 3 Medium | Stable Audio 2.5 | ACE-Step 1.5 XL | |
|---|---|---|---|
| Price | $0.035 per request | $0.68 per request | $0.00025 per second |
| A 30-second clip costs | $0.035 | $0.68 | $0.0075 |
| Length | 1 to 380 seconds | 1 to 190 seconds | 10 to 240 seconds |
| Sung lyrics | No | No | Yes |
| Sound effects | Yes | Yes | No (music) |
| Formats | MP3, WAV, FLAC, Ogg | MP3, WAV | FLAC, WAV, Ogg, MP3 |
How we ran it
Each model got the same prompts, one shot each, with the same seed (20261024) and its default settings, through POST /v1/audio/generations on October 24, 2026. ACE-Step 1.5 XL ran the beat as an instrumental ("lyrics": "[instrumental]") and skipped the sound effect, which is not what it is for. No result was retried or picked.
Prompt 1: a 30-second lo-fi beat
Warm lo-fi hip hop beat with soft Rhodes piano, vinyl crackle, mellow bass and brushed drums, 85 BPM
Stable Audio 3 Medium
Stable Audio 2.5
ACE-Step 1.5 XL
Prompt 2: a 10-second sound effect
A steam train passing through a station: approaching chuffing, a long whistle, brakes squealing, then fading away
Stable Audio 3 Medium
Stable Audio 2.5
When to use each
- Stable Audio 3 Medium for instrumental tracks and sound effects, especially long ones: a six-minute track costs what a ten-second one does.
- Stable Audio 2.5 when you already build on it and clips up to 190 seconds are enough.
- ACE-Step 1.5 XL for songs with sung lyrics, priced by the second.
Generate them on EmpirioLabs
curl https://api.empiriolabs.ai/v1/audio/generations \
-H "Authorization: Bearer $EMPIRIOLABS_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "stable-audio-3-medium",
"prompt": "Warm lo-fi hip hop beat with soft Rhodes piano, vinyl crackle, mellow bass and brushed drums, 85 BPM",
"duration": 30
}'The response is a job_id to poll at GET /v1/jobs/{job_id}. Change model to stable-audio-2-5 or ace-step-1-5-xl to run the same prompt on another model; each model page lists its own controls.
Frequently asked questions
Which one makes the longest clips?
Stable Audio 3 Medium, up to 380 seconds in one request.
Which one sings lyrics?
ACE-Step 1.5 XL. Both Stable Audio models make instrumental music and sound effects.
Does a longer clip cost more?
Not on the Stable Audio models, which are priced per request. ACE-Step 1.5 XL is priced per second of audio.
Try it
Playground | Models | Pricing
Stable Audio 3 Medium and Stable Audio 2.5 are Stability AI models. Powered by Stability AI.
