ComfyUI Node
Qwen3-TTS Custom Voice ⚡
Generate speech with a Qwen3-TTS CustomVoice model's built-in named speakers.
Qwen3-TTS Custom Voice ⚡
- qwen_tts_model
- AUDIO
◄textHello world►
◄speakerDylan►
◄languageAuto►
◄seed0►
◄max_new_tokens2048►
◄top_p0.80►
◄top_k20►
◄temperature1.00►
◄repetition_penalty1.05►
◄instruct►
◄custom_speaker_name►
◄unload_model_after_generatefalse►
Category🤖 CCTech/Qwen TTS
Inputs (13)
| Name | Type | Default | Description |
|---|---|---|---|
| qwen_tts_model | QWEN_TTS_MODEL | — | |
| text | STRING | Hello world | — |
| speaker | COMBO | Dylan | 9 options: Vivian, Serena, Uncle_Fu, Dylan, Eric, Ryan, +3 |
| language | COMBO | Auto | 11 options: Auto, Chinese, English, Japanese, Korean, German, +5 |
| seed | INT | 00–18446744073709550000 | — |
| max_new_tokens | INT | 20481–32768 | — |
| top_p | FLOAT | 0.800–1 | — |
| top_k | INT | 200–500 | — |
| temperature | FLOAT | 1.000–2 | — |
| repetition_penalty | FLOAT | 1.050.5–5 | — |
| instructopt | STRING | Voice-design instruction. Ignored by 0.6B CustomVoice checkpoints (upstream behavior). | |
| custom_speaker_nameopt | STRING | Overrides speaker with any string - for a fine-tuned checkpoint's own speaker names. | |
| unload_model_after_generateopt | BOOLEAN | false | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| AUDIO | AUDIO | — |