Qwen 3 TTS in ComfyUI – Local AI That Handles Long Dialogs with Emotion, Accent & Multi-Language (Patreon)

Tutorial Video : https://youtu.be/AR6JcdzbeW0

In this video, we dive deep into Qwen 3 TTS—the latest open-source AI text-to-speech breakthrough from the Qwen team—and show you how to run it locally in ComfyUI with full control. You’ll learn how to install the custom nodes, download models correctly (including the often-missed tokenizer and Voice Design components), troubleshoot common issues like SoX dependencies, and generate lifelike, multi-language, multi-character voiceovers—all on your own PC. Whether you're cloning voices from 3-second samples or designing custom narrators with emotional depth, Qwen 3 TTS delivers studio-quality audio with just 97ms latency and under 1.24% word error rate.

This tutorial is perfect for AI creators, indie developers, content producers, and ComfyUI power users who want professional-grade voice generation without relying on cloud APIs or paid services. If you’ve worked with WAN 2.2, LTX, or Sound-to-Video pipelines—or if you’re building audiobooks, animated shorts, or AI-driven storytelling—you’ll benefit from integrating Qwen 3 TTS into your local workflow. No subscription, no usage limits—just high-fidelity, controllable speech that supports 10 languages and maintains voice identity even during code-switching.

Why does this matter? Because voice is the next frontier of personalized AI content. With Qwen 3 TTS, you’re not just generating speech—you’re crafting characters with age, accent, emotion, and cultural nuance. From a 72-year-old British actor to a bilingual Shanghai nurse, this model lets you build rich vocal identities that stay consistent across long-form dialogue. And when paired with future video-generation tools, it becomes the foundation for fully synthetic, emotionally resonant media—made entirely offline.

Qwen3-TTS

https://github.com/QwenLM/Qwen3-TTS

ComfyUI-Qwen-TTS

https://github.com/flybirdxx/ComfyUI-Qwen-TTS

Remark:

In ComfyUI/models/qwen-tts/ folder

CMD , and activate your env. (Conda or python env)

Download the model file that you need.

For Qwen3-TTS 1.7B

hf download Qwen/Qwen3-TTS-12Hz-1.7B-Base --local-dir ./Qwen3-TTS-12Hz-1.7B-Base

hf download Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice --local-dir ./Qwen3-TTS-12Hz-1.7B-CustomVoice

hf download Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign --local-dir ./Qwen3-TTS-12Hz-1.7B-VoiceDesign

For Qwen3-TTS 0.6B

hf download Qwen/Qwen3-TTS-12Hz-0.6B-Base --local-dir ./Qwen3-TTS-12Hz-0.6B-Base

hf download Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice --local-dir ./Qwen3-TTS-12Hz-0.6B-CustomVoice

For Qwen3-TTS-Tokenizer-12Hz

hf download Qwen/Qwen3-TTS-Tokenizer-12Hz --local-dir ./Qwen3-TTS-Tokenizer-12Hz

About SoX - Sound eXchange Files

If you are missing 'SOX' lib or showing error in ComfyUI bootup time.

Go to

https://sourceforge.net/projects/sox/files/sox/

To fix this in your Conda environment for 2026, follow these steps:

1. Install the SoX Binary via Conda

You must install the actual binary (the .exe file) so that the command line can find it. Use the conda-forge channel:

bash

conda install -c conda-forge sox

Use code with caution.

2. Verify the Installation

Once the Conda installation finishes, restart your terminal or re-activate your environment, then run:

bash

sox --version

Use code with caution.

This should now return the version number (e.g., SoX v14.4.2) instead of an "internal or external command" error.

Attached Example Workflow for Dialog Inferencing