ByteDanceSeed Audio
Seed Audio 1.0
Seed Audio 1.0 combines a language model for semantic control with a diffusion model for acoustic generation. It covers speech, sound effects and ambient audio.
Technical guide
The API supports timing control and optional word timestamps. For image-conditioned voice generation, the text prompt contains the words to synthesize. Audio-reference mode instead supports a description of the desired sound.Seed Audio 1.0 HTTP API
Practical use
Practical guidance. For a sound scene, describe each sound and its timing. For narration, specify the exact words and listen for omissions. These are different acceptance checks even though the same model produces both.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | API |
| Model identifiers | seed-audio-1.0 |
| API output | Up to 120 seconds through a non-streaming HTTP API.Seed Audio 1.0 HTTP API |
| References | Up to three audio clips or one image; the API does not allow mixed image and audio references.Seed Audio 1.0 HTTP API |
Related models
- Seed TTS 2.0
ByteDance ยท Seed Speech
Context-aware speech synthesis model with text controls for tone, emotion and pace.
Launch date not confirmedAPI