Seed Audio

Mode: audio · Models: 2

Vendor: ByteDance Seed · Official API docs: ByteDance Seed

Seed Audio provides natural text-to-speech with named voices, voice cloning from reference audio, adjustable speech rate, loudness, and pitch, and WAV, MP3, PCM, or Ogg Opus output. The multilingual model expands synthesis to 20 languages.

Models

id	Name	Input type
seed-audio-1.0-multilingual	Seed Audio Multilingual	tts
seed-audio-1.0	Seed Audio	tts

CLI

# English or Chinese speech with a named voice
gen-ai generate -m seed-audio-1.0 \
  -p "Welcome to Picsart AI Playground." \
  --voice en_male_tim_uranus_bigtts --format mp3

# multilingual speech with optional voice-cloning references
gen-ai generate -m seed-audio-1.0-multilingual \
  -p "Bonjour et bienvenue." --audio-urls ./voice-reference.wav

Run gen-ai models info <model-id> --json for the full voice catalog: 95 voices on Seed Audio and 185 on Seed Audio Multilingual.

MCP

{ "name": "picsart_generate",
  "arguments": {
    "model": "seed-audio-1.0-multilingual",
    "prompt": "Bonjour et bienvenue.",
    "voiceId": "en_male_tim_uranus_bigtts",
    "format": "mp3"
  } }

Parameters

Full parameter surface for both models, sourced from gen-ai models info <id> --json.

seed-audio-1.0-multilingual — Seed Audio Multilingual

Try seed-audio-1.0-multilingual in Playground ↗

Input type: tts

The multilingual model accepts the same parameters as seed-audio-1.0, with a 185-voice catalog spanning 20 languages.

seed-audio-1.0 — Seed Audio

Try seed-audio-1.0 in Playground ↗

Input type: tts

Param	CLI flag	Type	Values
prompt	-p	text	required
voiceId	--voice	enum	95 named voices (default en_male_tim_uranus_bigtts)
audioUrls	--audio-urls	file	reference audio (up to 3)
imageUrls	-i	file	reference image (up to 1)
format	--format	enum	wav · mp3 · pcm · ogg_opus (default wav)
sampleRate	--sample-rate	enum	8000 · 16000 · 24000 · 32000 · 44100 · 48000 (default 44100)
speechRate	--speech-rate	range	-50–100 (default 0)
loudnessRate	--loudness-rate	range	-50–100 (default 0)
pitchRate	--pitch-rate	range	-12–12 (default 0)
aigcWatermark	--aigc-watermark	boolean	true · false (default false)

Pricing

gen-ai pricing seed-audio-1.0-multilingual

Text-to-speech cost is resolved from the selected model and generation parameters before execution.