Audio Sandbox.
A third-person 3D playground where every sound effect is generated by Stable Audio 3 as you play.

Project Details

@maxfieldhulkerSAI
Builder since 2026
- Published
- 11/05/2026
- Model
- Stable Audio 3.0 Large
- Hosting
- API
- License
- MIT
Endpoints Used
- Text to Audio
What It Does
You walk a character around a grass field with nine surface pads (concrete, gravel, ice, metal, moss, sand, slate, snow, water), a small indoor room with wood and linoleum floors, some walls to throw at (brick, plaster, sheet metal, glass, a plank fence), a bag with a ball and a phone in it, and a skee-ball machine. Every footstep, jump, landing, ball bounce, score chime, menu click and the ambience bed is a clip generated by Stable Audio 3 from a prompt template. On first load the game generates the whole sound set, nearest things first, then plays from cache. Stand on a surface, next to a wall, or at the machine and press E to open a tuning card: describe what it should sound like, regenerate, audition the takes, pick one. The phone in your bag opens Stable Stack, a merge puzzle game whose sounds were made the same way.
Code Snippets
curl -X POST https://api.stability.ai/v2beta/audio/stable-audio/text-to-audio \ -H "Authorization: Bearer $STABILITY_API_KEY" \ -F model=stable-audio-3 \ -F prompt="A punchy lo-fi drum loop with dusty vinyl crackle, 90 BPM." \ -F duration=30 \ -o result.mp3
How It Uses Stable Audio
Text to Audio
Generates every sound effect. One call per take, with model=stable-audio-3, duration from 1 s (footsteps) to 30 s (the ambience bed), steps 8 and cfg_scale 7 by default (both adjustable in Settings), a random seed, and output_format mp3 or wav.



