Crowd size sets the sound of applause. Four people clapping in an office and four thousand in an arena are different sounds: in the large crowd the individual claps merge into noise, and in the small one you can count the hands. Choosing the wrong size makes a video sound canned, and stock clips rarely let you control it.
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.
six people applauding politely in a small meeting room, individual claps clearly audible, close and dry
theatre audience of several hundred applauding warmly, dense overlapping claps, wooden hall reverb
stadium crowd of thousands applauding and cheering, applause merging into a continuous roar, vast open space
single person slow clapping sarcastically, deliberate spaced claps with a hard room echo, empty hall
audience applause starting scattered and building into a full standing ovation with whistles, large hall
conference audience applauding politely at the end of a talk, even and restrained, medium hall with air conditioning
applause tapering off and dying away to a few last claps, then room tone, medium hall
small polite golf gallery clapping quietly, soft muffled claps outdoors with birdsong behind
small comedy club audience laughing then breaking into applause, close intimate room, low ceiling
one sharp hand clap in an empty room, bright transient with a clear slap-back echo, dry and close
These models respond to physical description. Name the material, the action and the space.
six people applauding politely in a small meeting room, individual claps clearly audible, close and dry
A number plus "individual claps clearly audible" keeps the model from defaulting to a stadium, which it does on the word applause alone.
stadium crowd of thousands applauding and cheering, applause merging into a continuous roar, vast open space
"Merging into a continuous roar" is the acoustic signature of a big crowd, and it separates size from volume.
audience applause starting scattered and building into a full standing ovation with whistles, large hall
Applause usually has an arc. Describing it gives you something to cut against picture instead of a flat bed.
single person slow clapping sarcastically, deliberate spaced claps with a hard room echo, empty hall
"Deliberate spaced" plus an empty room makes the joke: the echo between claps is what makes it sound sarcastic.
As a crowd grows, the gaps between claps fill in until the sound is continuous. That transition happens at around a hundred people, and it is the cue an audience uses. When applause sounds like the wrong size, change the density words and leave the level alone: "individual claps clearly audible" for small, "dense overlapping" for medium, "merging into a continuous roar" for large. A stadium turned down sounds like a distant stadium.
Applause that starts at full volume and stops dead sounds like a stock clip. Real applause ramps: a few hands, then the room joining, then the peak, then a ragged tail as people stop at different times. Ask for the shape you need, and generate the tail as its own clip if you want to control where the scene lands. The fade-out sample in this collection is for that.
Dry, close, mid-sized applause with no room in it sounds like 1980s television, because that is what a laugh-track machine produced. For something contemporary, give it a specific space (a wooden hall, a low-ceilinged club, an outdoor gallery) and let the reflections place it. It costs a few words in the prompt.
The models run on your machine, which changes the cost, the privacy and what you can automate.
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.
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.
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}'