Seed Audio 1.0. Score a whole scene in one pass
By Steven Van ·
Seed Audio 1.0 from ByteDance generates speech, sound effects, and music together from a single prompt, so a clip gets voice, ambience, and a soundtrack…
Higgsfield announced Seed Audio 1.0. Score a whole scene in one pass on June 26, 2026.
What changed
Seed Audio 1.0 from ByteDance generates speech, sound effects, and music together from a single prompt, so a clip gets voice, ambience, and a soundtrack in one step instead of three tools and a manual mix. Multi-speaker with continuity. Several voices hold across the scene, with emotion and accent control. Ambience, music, and Foley. Rain, traffic, room tone, a musical bed, footsteps, and textures. Context-aware. Give it text, an image, or an audio reference to guide the result. To get started: Audio → model selector → Seed Audio. Describe the scene, including how long each element should run. Availability: outputs in WAV, MP3, PCM, or OGG. Voice cloning supports up to 3 audio references. Length is set through your prompt, not a duration field.
