Six days after Seedance 2.5, ByteDance opened enterprise-access applications for Seed Audio 1.0 — a non-streaming text-to-speech model that generates voice, music, and sound effects together in a single pass, rather than routing each through a separate tool. Alongside natural speech and a bank of preset voices, it offers reference-audio guidance to clone a voice or style, image-guided audio that reads a still frame to produce matching ambience and score, and fine-grained control over speed, volume, pitch, and output format. For AI filmmakers, the pitch is a unified soundstage: instead of stitching a TTS narrator, a music generator, and a sound-effects library into one timeline, a single model can voice the dialogue, underscore the scene, and lay in the room tone at once — the audio counterpart to the multimodal generation ByteDance is pushing on the video side. Early testers on X flagged the obvious gap — no duration controls yet — but praised the ambience and lifelike speech. Launched as an enterprise beta through BytePlus with no public pricing.