A Commodore 64 SID chip, a Sega Genesis YM2612 FM synth and a Yamaha OPL2 AdLib card are three different pieces of hardware with three different timbres. A prompt that says only "chiptune" leaves the model to guess which one you mean. Name the chip and the voice it is known for (filtered square wave, FM bell, tracker sample) and Stable Audio renders a loop in that machine's sound.
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.
Commodore 64 SID chip chiptune loop, MOS 6581 filtered square wave lead synth arpeggio, pulse-width modulation bass synth, gritty analog filter sweep, upbeat major key, 130 BPM, seamless loop
Sega Genesis YM2612 FM synth action theme, punchy FM bell lead synth melody, driving FM bass synth, crisp PSG noise-channel hi-hats, energetic 16-bit action game score, 140 BPM, seamless loop
NES 2A03 chiptune dungeon theme, triangle wave bass synth ostinato, pulse wave lead synth melody with duty-cycle changes, noise-channel percussion, moody minor key, seamless loop
Yamaha OPL2 AdLib FM synthesis boss battle theme, aggressive FM brass stab synth hits, driving FM bass synth line, retro PC soundcard timbre, tense minor key, 145 BPM, seamless loop
Amiga ProTracker module overworld theme, sampled square wave lead synth melody, plucked chiptune bass synth, tracker-style drum breaks, bright cheerful major key, seamless loop
8-bit chiptune hero anthem, bright square wave lead synth melody, driving triangle wave bass synth, punchy noise-channel drums, energetic major key, gang-vocal chant
These models respond to physical description. Name the material, the action and the space.
Commodore 64 SID chip loop, MOS 6581 filtered square wave lead synth, pulse-width modulation bass synth, analog filter sweep
"Filtered" and "pulse-width modulation" separate a SID from a generic square wave. The SID's analog filter and PWM bass are its signature.
Sega Genesis YM2612 FM synth theme, FM bell lead synth, FM bass synth, PSG noise-channel hi-hats
Use "FM" plus an instrument word (bell, bass). FM synthesis has a metallic, bell-like character that square-wave chiptune lacks, and naming the synthesis method gets it.
NES 2A03 chiptune theme, triangle wave bass synth, pulse wave lead synth with duty-cycle changes, noise-channel percussion
The 2A03 has four voices: two pulse, one triangle, one noise. Name each by its role (bass, lead, percussion) to keep the arrangement authentic. Otherwise you get a wall of square waves.
Yamaha OPL2 AdLib FM synthesis theme, FM brass stab synth hits, FM bass synth line, retro PC soundcard timbre
OPL2 is a different FM chip from the Genesis's YM2612. "AdLib" and "soundcard" push the model toward its thinner, brassier late-80s PC voice.
Amiga ProTracker module theme, sampled square wave lead synth melody, plucked bass synth, tracker-style drum breaks
The Amiga played back samples instead of synthesising its voices. "Tracker" and "sampled" push the output toward that grittier, sample-based texture.
8-bit chiptune hero anthem, bright square wave lead synth melody, driving triangle wave bass synth, punchy noise-channel drums, energetic major key, gang-vocal chant
ACE-Step's `prompt` sets genre and instrumentation, as on the instrumental loops above. A separate `lyrics` field ("[verse]\nRise up, hero\nThe quest begins") gives it words to sing over the beat.
"Retro" and "old-school" describe an era. "SID", "YM2612", "2A03" and "OPL2" describe a sound. Each chip has a well-documented timbre: the SID's analog filter, the YM2612's FM bell tones, the 2A03's four fixed voices, the OPL2's thinner AdLib character. Naming the chip tells the model which of these you want.
Stable Audio 3 SFX is tuned for one-shot sound design: a door, a footstep, a laser. A 15-second chiptune loop needs a lead melody, a bass line and percussion that stay in time and in key for the whole length. That is a composition task, and a music model (Stable Audio 3 Small Music or ACE-Step) holds it together better. Use one whenever the result should sound like an instrument performance.
These models can render a loop where the last beat lines up with the first. Ask for it explicitly and write the prompt as a fixed-tempo phrase ("130 BPM", "seamless loop"). If the loop point still clicks, trim a few milliseconds off the tail or add a short crossfade in an editor.
The Stable Audio loops above have no voice, because Stable Audio 3 Small SFX has no vocal path. It renders instruments only. ACE-Step is a separate model family built for lyrics-conditioned vocal music: give it a `lyrics` block and it sings those words over the arrangement described in `prompt`. Use it when you need a hook, a chant or a sung line, such as a hero anthem or a victory chorus.
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}'