A single request returns one audio clip. Parallelize multiple requests on the client to get multiple variants. The response body is a raw audio/wav byte stream (48 kHz, 16-bit, stereo).
Request parameters body.args
Field
Type
Description
prompt*
string
Text prompt. English or Chinese, any length. The more specific, the more controllable.
cfg
float
Classifier-Free Guidance. 0–10. Higher values follow the prompt more closely but distort more easily. default: 4.5
seed
int32
Random seed, for reproducibility. Omit to randomize each call.
n
int (1–4)
Number of variants per request. n=1 returns an audio/flac stream; n≥2 returns application/json{count, sample_rate, format, audios:[base64...]}. default: 1
05
Errors
All error responses are application/json with shape
{ code, message, request_id }.
400
BAD_REQUEST
JSON Schema validation failed — usually a typo or missing field such as args.prompt.
401
UNAUTHORIZED
Authorization header missing or token expired.
413
PAYLOAD_TOO_LARGE
Request body too large. See §01.
422
VALIDATION_ERROR
Parameters are the right type but out of range — e.g. cfg = 12.
429
RATE_LIMITED
Exceeded 10 req/min. Response carries a Retry-After header.
503
QUEUE_FULL
Inference queue saturated. Client should retry with exponential backoff.
06
AI Integration — Copy-to-Prompt
Drop a battle-tested prompt into Claude Code, Cursor, or Codex and ship the integration in under a minute. Covers the API contract, auth, retries, and error handling — plus TypeScript / Python scaffolding.
AI-READY PROMPT
— tokens
Integrate fast with AI
Tested on major coding agents. Contains API shape, auth, retries, error handling, and example I/O. Just paste and say "do this in my stack".
07
Sound Effects Plaza
Community samples, publicly shared. Every prompt can be adapted.
S
SenseAudio · 音频智能
Give every creator the power to sound-design in one sentence.
Sign in to call the diffusion sound-effects model — sub-second latency, library-grade stereo.
DEMO · Secure connection
Welcome back to SenseAudio
Sign in to start calling /v1/audio/sound_effect_generate.