Ambience and background sound effects
For a background bed, settle the length first and the content second. Check the loop length your engine or edit needs.
Listen: 11 samples from the ambience collection
Every clip shows the exact prompt that produced it. The render script reads its prompts from the same file as this page. Copy one, change a word and generate your own.
Tavern folk music (loop)
via audiocpp-ace-stepcozy medieval tavern folk music loop, acoustic strummed instrument, warm mellow tone, relaxed tempo, seamless loop
Airport lounge
via audiocpp-stable-audio-3-sfxairport departure lounge, distant tannoy announcements, rolling cases and quiet conversation, large hall
Quiet library
via audiocpp-stable-audio-3-sfxquiet library room tone, distant page turns, a single cough and faint chair movement, soft reverb
Subway platform
via audiocpp-stable-audio-3-sfxunderground subway platform, tiled reverb with distant train rumble and echoing announcements
Hospital corridor
via audiocpp-stable-audio-3-sfxhospital corridor at night, distant monitor beeps, footsteps and a low ventilation hum, hard reverb
Medieval market
via audiocpp-stable-audio-3-sfxmedieval market square, wooden carts, livestock, hammering and layered period chatter, open air
Beach day
via audiocpp-stable-audio-3-sfxsunny beach ambience, waves breaking with distant children playing and gulls overhead, wide open
Mountain summit
via audiocpp-stable-audio-3-sfxexposed mountain summit, thin steady wind with faint distant birdsong and no traffic, vast and empty
Server room
via audiocpp-stable-audio-3-sfxdata centre server room, dense layered fan noise with a high electrical whine, enclosed and constant
Old attic
via audiocpp-stable-audio-3-sfxold house attic room tone, timbers creaking, faint wind through gaps and distant traffic, enclosed
Desert night
via audiocpp-stable-audio-3-sfxdesert at night, faint wind over sand with occasional distant insect chirps, vast silence, seamless loop
Prompts that work
These models respond to physical description. Name the material, the action and the space.
Office room tone
quiet office room tone, faint hum of fluorescent lights and distant HVAC, no voices, static and dry
Open-ended ambience prompts tend to include indistinct speech, which you rarely want under dialogue. Say "no voices".
Cafe bed
busy cafe ambience, overlapping indistinct chatter, occasional cup clinks, steady bed, 3 seconds
"Overlapping indistinct chatter" avoids intelligible words, so the crowd bed doesn't produce distracting speech.
Night forest
quiet forest at night, steady cricket chorus, faint distant owl, outdoor, 3 seconds
One steady element (crickets) plus one occasional element (a faint owl) gives the bed texture without turning it into a sequence of foreground events.
Sci-fi interior
sci-fi spaceship interior ambience, low mechanical hum, faint electronic beeps, sealed pressurized feel, 3 seconds
"Sealed pressurized feel" describes the space. The model shapes the whole bed as if heard inside an enclosed hull.
Loopable dungeon bed
dark stone dungeon ambience loop, distant dripping water, low reverberant drone, no voices, 8 seconds, seamless loop
"Seamless loop" makes the model avoid a strong attack or decay at the start and end, so the bed loops without clicking at the seam.
Making ambience sounds that fit
Set the duration first
Before writing what a bed contains, check the length your engine or edit needs. A two-second sting behind a menu screen and an eight-second seamless loop behind a level are different prompts. State the target length and, if it needs to loop, say "seamless loop" so the model avoids a hard attack or a fading tail that would click at the seam. A bed at the wrong length is unusable however good it sounds.
Layer one steady element with one occasional one
Convincing ambience beds have a steady floor (hum, hiss, a chorus of crickets) plus one or two occasional events that break it up (a distant owl, a car passing, a cup clink). Name both in the same prompt and mark the occasional element as "faint" or "distant". The bed then sounds neither flat nor too busy to sit under dialogue.
Exclude speech unless you want it
Crowd and cafe ambiences often drift toward intelligible or semi-intelligible speech unless you tell the model otherwise. Say "indistinct chatter" for crowd texture that doesn't pull focus, or "no voices" for anything that has to sit under dialogue.
Use a music model for a music bed
Every other sample on this page is environmental sound (hums, chatter, wind, dripping water), and Stable Audio 3 SFX handles all of it. A tavern with background music playing is a composition task: an instrument performing a melody that holds together for the whole loop. Use ACE-Step for that, with `lyrics` left empty for an instrumental render, as on the chiptune-music pages.
Why generate locally
The models run on your machine, which changes the cost, the privacy and what you can automate.
Offline after download
Weights are fetched once. After that, generation, the library and the API all work with the network off. Everything lives in one SQLite file and one folder of WAVs under ~/.soundbox, and nothing is uploaded.
Your agent can use it
Soundbox runs an MCP server in the same process, with tools for generating and listing sounds. Generated files come back as absolute paths, so an agent can copy them into a project.
Licences stated up front
Every model's terms are shown before you download it, and non-commercial weights need an explicit confirmation. Soundbox is MIT. The licence of the model that made a sound governs what you may do with it.
claude mcp add --transport http soundbox http://127.0.0.1:17494/mcpcurl -X POST http://127.0.0.1:17494/api/generate \
-H 'Content-Type: application/json' \
-d '{"model":"audiocpp-stable-audio-3-sfx","prompt":"8-bit coin pickup blip","duration_s":1}'Frequently asked questions
- Can I use these sounds in a commercial video or game?
- It depends on the model that made them. Soundbox shows each model's licence before you download its weights. The Stable Audio 3 models use the Stability AI Community License: commercial use is free below USD $1M annual revenue, and above that Stability requires an enterprise licence. ACE-Step 1.5 is MIT. MiniMax's models allow commercial use under a revenue cap and require an attribution notice. Some voice and stem-separation models are non-commercial only. Soundbox itself is MIT, but the audio is governed by the licence of the model that made it. Read that licence before you ship; this summary is not legal advice.
- Does this run offline?
- Yes, once the weights are downloaded. Soundbox runs the models locally and stores everything in one SQLite file and one folder of WAVs under ~/.soundbox. There is no account, API key or upload step. Only the initial model download needs the network.
- What hardware do I need?
- An NVIDIA GPU with 8 GB or more. CPU-only is supported, but each generation then takes minutes. That is fine for a few short clips and slow for a large batch.
- Can my coding agent generate these for me?
- Yes. While Soundbox is running it serves MCP at http://127.0.0.1:17494/mcp, with tools for generating and listing sounds. An agent can generate a sound as part of the task it is already doing and gets back an absolute file path to use in your project.
- How is this different from downloading a sample pack?
- A pack gives you a bed at a fixed length that you have to crossfade-loop or trim, with whatever content the pack included. Generating lets you set the duration your engine loops at and the mix of steady floor and occasional events your scene needs, including whether there is speech. It costs time and a GPU.