stable-audio

Generate rights-safe music and sound effects by using text-to-audio workflows with documented provenance, rights-cleared source uploads, and final audio quality checks to ensure production readiness. This prevents unsuitable speech workflows and rights risks.

Audio inpainting works by applying audio-to-audio transformations with appropriate source-audio strength and prompts to regenerate or fill missing sections of existing loops, foley, or ambience while preserving the original motif and continuity.

Stable Audio 3 endpoints support long duration sound beds through asynchronous job polling, allowing you to generate extended instrumental beds, ambience, and sonic branding assets while managing latency and parameter limits.

Choose between hosted and local open-weight audio generation models by evaluating duration, latency, quality, and local-data requirements to route your text-to-audio, inpainting, and continuation workflows to the appropriate endpoint.

Limitations for uploading source audio include strict rights-cleared content requirements and provenance documentation, preventing rights-uncleared uploads and unsuitable speech workflows from entering the audio transformation pipeline.

Generated music beds have artifacts and fail audio quality checks due to improper generation steps, seeds, or parameter limits, requiring a final review for loudness, continuity, editability, and delivery readiness before media project integration.