Seed Audio 1.0

Seed Audio 1.0 is ByteDance Seed's multimodal AI audio generation model for building complete sound scenes from text, image, and audio inputs. In one workflow, it can combine multi-speaker dialogue, expressive voices, native accents, ambience, background music, and foley-style sound effects instead of producing speech alone.

Prompts can describe speakers, timing, emotion, language, location, music direction, and sound events together. This makes the model useful for cinematic scenes, localized dialogue, ads, podcasts, games, and other projects where voice and the surrounding soundscape need to work as one.

Inputs: Text, image, and up to 3 audio references; Sound layers: Dialogue, voice, ambience, music, and sound effects; Output: One coordinated audio scene up to 2 minutes.

How to use: write a sound-scene prompt, optionally add an audio or image reference, choose output settings, then generate and review the audio in the SeedAudio.co workspace.

Use cases

• Scene Sound & Foley for short dramas, food videos, and AI video post-production.
• AI Music to produce original soundtracks and background music for films, ads, and regional formats.