### Audio — pick this default
1. **Default audio generation (text-to-audio, voice-style, SFX, ambience, foley, impacts, environmental audio, or music-like audio)** → Seed Audio 1.0 (`seed_audio`). Requires `--prompt`; use optional `--audio-references`/`--image-references` only when references are provided.
2. **Create music, backing tracks, jingles, or instrumental beds with the specialist legacy music model** → Sonilo Music (`sonilo_music`) only when the user names Sonilo or Seed Audio is not appropriate. Requires `--prompt` and `--duration`.
3. **Create SFX with the specialist legacy SFX model** → Mirelo Text to Audio (`mirelo_text_to_audio`) only when the user names Mirelo or Seed Audio is not appropriate. Requires `--prompt` and `--duration`.
4. **Add soundtrack/audio to a generated video ad** → use Marketing Studio Video with `--generate_audio true`, not Seed Audio/Sonilo/Mirelo.
flowchart TD
A[音訊需求] --> B{為影片廣告加聲音?}
B -->|是| MS[Marketing Studio Video<br>--generate_audio true]
B -->|否| C{類型?}
C -->|旁白/對白 TTS| T[Eleven v3 / MiniMax Speech 2.8 HD / Seed Speech / VibeVoice / Inworld / Qwen Audio]
C -->|SFX/環境/擬音/類音樂| SA[Seed Audio 1.0（預設）]
C -->|指名音樂模型| SO[Sonilo Music]
C -->|指名 SFX 模型| MI[Mirelo Text to Audio]
C -->|對嘴/換聲/配音| LS[Lipsync、voice-change、dubbing 工作流]