Authors and ModelsStepFun

StepAudio 3 Gen

StepAudio 3 Gen
StepFun · Audio Generation
POST /v1/audio/generations

Generates dialogue, sound effects, ambience, and background music together in one clip from a written scene.

At a glance

FieldValue
Model idstepaudio-3-gen
Model release date2026-09-09
Input modalitiesText
Output modalitiesAudio
Context window-
Weight precision-
RegionInternational
Featuresaudio_generation, text_to_speech, sound_effects, music_generation, voice_control
Native inferenceNo
NewYes
Supported endpointsPOST /v1/audio/generations
Alternate model idsstepaudio-3-gen-preview, stepfun/stepaudio-3-gen

Pricing

ChargeSpecRate
Generationper 10,000 characters$0.36

Example request

curl https://api.empiriolabs.ai/v1/audio/generations \
  -H 'Authorization: Bearer $EMPIRIOLABS_API_KEY' \
  -H 'Content-Type: application/json' \
  -d '{"model": "stepaudio-3-gen", "prompt": "warm jazz piano", "duration": 8}'

Parameters

ParameterTypeRequiredDefaultDescription
promptstringno-The scene to generate. Wrap a sound effect or music cue in square brackets and a delivery direction in parentheses.
instructionstringno-Global guidance for the setting, background music, and emotional tone. Maximum 500 characters. · Max: 500
rolesstringno-Optional speakers, as objects with a name and a voice description. All names and descriptions together are capped at 500 characters.
scriptsstringno-Optional ordered lines, as objects with a speaker and text. Capped at 1,000 characters in total. Use this instead of prompt for multi-speaker scenes.
response_formatenumno"mp3"Output audio format. · Allowed: mp3, wav, flac, opus, pcm
sample_rateenumno24000Output sample rate in Hz. · Allowed: 8000, 16000, 22050, 24000, 48000
speednumberno1Speech speed. · Range: 0.5 – 2
volumenumberno1Output volume. · Range: 0.1 – 2
text_normalizationenumno"standard"How numbers, dates, and symbols are read aloud. · Allowed: standard, enhanced

Notes

Describe a scene and the model performs it as one finished clip: spoken lines, sound effects, ambience, and background music together. Send a prompt, or use roles and scripts for multi-speaker control. Wrap a sound effect or music cue in square brackets and a delivery direction in parentheses. Roles and instruction are capped at 500 characters each and scripts at 1,000. Output formats are mp3, wav, flac, opus, and pcm. Reference voices and voice cloning are not available on this model. This model is in preview and its capabilities may change.

Machine-readable schema: GET https://api.empiriolabs.ai/v1/models/stepaudio-3-gen.