ByteDance's Seed launches audio creation model Seed Audio 1.0

ByteDance's Seed has launched Seed Audio 1.0, an audio creation model that provides unified modeling of human voices, sound effects, and ambient sounds for complete sound scenarios, enabling end-to-end film-quality audio creation. The model supports unified arrangement of multiple elements and 100ms precision timing control, maintaining consistent timbre over extended periods based on reference audio, and naturally generating audio in over 20 languages. Evaluations show that audio usability exceeds 90% in most scenarios.