声音风格指引 (QwenTTS)

If you're generating Chinese speech with Qwen3-TTS, this is the node you didn't know you wanted. 声音风格指引 (QwenTTS) - "voice style guide" - is the Chinese-language sibling of Voice Instruct (QwenTTS): the same utility idea, but its presets are written in Chinese and aimed at Chinese output, because that's where the model listens best.

What it does

Like its English sibling, this is a pure utility node. No model loads, no audio comes out - it assembles a voice-style instruction string from two dropdowns and hands it to your generation node. The dropdowns are even labeled in Chinese, which is the tell that this node is built for Chinese-language pipelines.

• 角色 (character) - 14 presets: 自动 (Auto), 女声 (female), 男声 (male), 年轻女声 (young female), 年轻男声 (young male), 小女孩 (girl), 小男孩 (boy), 童声 (child), 青少年 (teen), 成年 (adult), 老年女声 (senior female), 老年男声 (senior male), plus a couple more.
• 风格 (style) - 47 presets: 温暖 (warm), 轻柔 (gentle), 平静 (calm), 愉快 (cheerful), 友好 (friendly), 严肃 (serious), 悲伤 (sad), 愤怒 (angry), 兴奋 (excited), 轻声 (soft), 低沉 (deep), and more.
• 自定义风格指引 (custom style guide) - optional free text that overrides both presets when filled in (the tooltip notes it has top priority).

Output is a single VOICE_INSTRUCT string, assembled from the selected character + style (or your custom text), ready to feed into the instruct input of a Custom Voice or Voice Design node. Its presets come from the pack's voice_instruct_zh.json, a separate file from the English one.

Why the separate node exists

It's not just translation. The Chinese presets are written to match how Qwen3-TTS handles Chinese prosody - which is better than English, honestly. Qwen3-TTS is at its strongest on Chinese output, and a well-worded Chinese instruction is the difference between a flat read and a believable character. The English node's presets are tuned for English phrasing; mixing them into Chinese generation is exactly the language mismatch that produces the "ignores everything" complaints.

Install

Nothing special - part of the same pack, same install:

cd ComfyUI/custom_nodes && git clone https://github.com/1038lab/ComfyUI-QwenTTS.git
python3 -m pip install -r ComfyUI/custom_nodes/ComfyUI-QwenTTS/requirements.txt

Or use ComfyUI Manager, search ComfyUI-QwenTTS. Restart, and it'll show up under the 🧪AILab/🎙️QwenTTS category. One rule of thumb for Chinese pipelines: set the generation node's language to Chinese (not Auto) and feed it this node's instruction - that combination is the pack's most reliable path to a good Chinese character voice.