Seed Audio 1.0

What is Seed Audio 1.0?

Seed Audio 1.0 treats a whole scene, not a single voice, as the thing it generates. A prompt of up to 3,000 characters can describe the speakers, their lines and mood, the room and the key sound effects, and the model returns one mixed track of up to two minutes. You steer the voice with a written description, reference clips, a saved BytePlus voice or an image.

It is ByteDance's current audio creation model, with no open weights; you use it through the Seed Speech API on BytePlus, which lists it as Audio 1.0. Most text-to-speech models voice only the words, while Seed Audio adds effects and ambience in the same pass.

Who is Seed Audio 1.0 best for?

Seed Audio 1.0 is best for video, game and podcast teams who want dialogue, sound effects and ambience generated together as one timed scene from a prompt, in more than 20 languages, through an API.

• Dubbing and re-voicing

  The prompt can set when each line starts, in 100-millisecond steps, so dialogue lands on cue against a video.

• Whole sound scenes

  Dialogue, sound effects and background ambience come out of one prompt already timed together, rather than as separate files to mix.

• Localized content

  It speaks more than 20 languages and can keep a character's voice recognizable when the language changes.

• Custom character voices

  Describe a voice in words, give an authorized sample or both, with no training run for each new speaker.

What's included with Seed Audio 1.0?

Scene generation, dialogue timing in 100-millisecond steps, voice and image references, word-level timestamps, several output formats and an optional AI watermark.

Scene generation

Speech, sound effects and ambience are generated together from one prompt, so they share timing and room sound.

Dialogue timing

Set in the prompt when each character's line enters, to the nearest 100 milliseconds.

Voice references

Use up to three reference clips of 30 seconds each, a saved TTS 2.0 or cloned BytePlus voice, or one reference image.

Word-level timestamps

Turn on subtitles and the response lists start and end times for every sentence and word.

Output settings

WAV, MP3, PCM or Ogg Opus at 8 to 48 kHz, with speed, volume and pitch controls.

Optional AI watermark

Add an audible marker at the end of the clip or a metadata tag that labels it as generated audio.

How much does Seed Audio 1.0 cost?

Seed Audio 1.0 costs $0.15 per minute of generated audio on ByteDance's BytePlus, billed by the second, with 60 free minutes when you activate it and prepaid packs from $28.50 for 200 minutes.

Prices are in US dollars. A 2-million-minute pack costs $240,000 ($0.12 a minute) and is also valid for 1 year. Billing counts the model's raw output length, before any speed change. BytePlus sells the model under Seed Speech, where its docs and pricing call it Audio 1.0. Full pricing on docs.byteplus.com. We checked these prices on 28 Sep 2026.

How do readers rate Seed Audio 1.0?

No one has rated Seed Audio 1.0 yet. Be the first.

Seed Audio 1.0 pros and cons

It builds a whole timed scene from one prompt at a simple per-minute price, but clips stop at two minutes, only dialogue timing is controllable, and it runs only on BytePlus.

What are the best Seed Audio 1.0 alternatives?

ElevenLabs for separate speech, effects and music models with a free plan, Gemini 3.1 Flash TTS for cheaper speech-only work, and Camb.ai for finished video dubs.

ElevenLabsFree

Choose it when you want speech in 90+ languages plus separate sound effects and music models under one plan, with a free tier and paid plans from $6 a month; you mix the layers yourself.

Gemini 3.1 Flash TTSFree

Choose it when you only need speech: it has a free tier, handles multi-speaker dialogue in 70+ languages, and paid output works out to about $0.03 a minute.

Camb.aiFree

Choose it when you want an existing video translated and dubbed for you rather than prompting each line; its Premier plan at $75 a month includes 16 minutes of dubbing.

Questions about Seed Audio 1.0