Actions18
- Image Actions
- Video Actions
- Audio Actions
- LLM Actions
- Creative Workflow Actions
- Model Actions
- Account Actions
Audio → Generate
AI-generatedSummary
Generate AI-composed audio tracks using a selected audio model and detailed prompts describing desired music characteristics, length, and style.
Inputs
- audioModelSearch — Optional text to filter the list of available audio models by name or tag.
- audioModelId (required) — ID of the audio model to use for generation, chosen from the filtered list or pasted manually.
- audioPositivePrompt (required) — Text description of the desired music or audio, including genre, mood, instruments, etc.
- audioNetwork — Choice of network type for generation speed and cost: 'Fast' (default) or 'Relaxed'.
- audioDuration — Length of the generated audio in seconds (minimum 10, maximum 600; default 30).
- audioAdditionalFields — Additional optional fields to control lyrics, language, BPM, time signature, key/scale, AI composer mode, prompt strength, creativity, sampler, scheduler, and other advanced music parameters.
- generationSettings — Optional settings like negative prompt, style prompt, number of audio tracks (1-10), inference steps, guidance, and random seed for generation control.
- output — Output options including whether to download audio as binary data (recommended) and audio format (MP3, FLAC, WAV).
- advanced — Advanced options for token type (Spark or SOGNI) and an explicit generation timeout in milliseconds.
Output shape
One or more generated audio tracks, returned either as binary data (if download enabled) or URLs.
The number of tracks generated defaults to one but can be set up to ten per request. Audio download is recommended to avoid expiration of temporary URLs. Duration must be between 10 and 600 seconds. Model selection dynamically updates based on the search filter. Optional parameters influence style, composition, and generation behavior, with sensible defaults if omitted.