ComfyUI_Dia
Dia TTS model capable of generating ultra-realistic dialogue in one pass. ComfyUI node.
Dia's ComfyUI Node
Text-to-Speech, Voice Cloning, Generating Ultra-Realistic Two-Person Conversations in One Go.
Supported oral tags include (laughs), (clears throat), (sighs), (gasps), (coughs), (singing), (sings), (mumbles), (beep), (groans), (sniffs), (claps), (screams), (inhales), (exhales), (applause), (burps), (humming), (sneezes), (chuckle), (whistles).
📣 Updates
[2025-05-29]⚒️: Speaker audio can be loaded and saved separately.
[2025-05-19]⚒️: Released v1.1.0. Text-to-speech generation of arbitrary length is possible (requires splitting with blank lines). Speakers can be saved and loaded directly afterward.
[2025-04-24]⚒️: Released v1.0.0.
Usage
- Prompt and text format must be as follows, use blank lines for particularly long ones:
[S1] Hi, how are you.
[S2] Fine, thank you, and you?
[S1] I'm fine, too.
[S2] What are you planning to do?
[S1] Hi, how are you.
[S2] Fine, thank you, and you?
-
Generating conversation by cloning voice:
-
Loading saved speaker:
-
Automatically generating conversation with Gemini:
https://github.com/user-attachments/assets/6b27114d-aa9e-4f70-99c1-683994621402
Installation
cd ComfyUI/custom_nodes
git clone https://github.com/billwuhao/ComfyUI_Dia.git
cd ComfyUI_Dia
pip install -r requirements.txt
# python_embeded
./python_embeded/python.exe -m pip install -r requirements.txt
Model Download
- Dia-1.6B: Download the entire directory and place it under the
ComfyUI/models/TTSdirectory. - weights.pth: Download and rename to
weights_44khz_8kbps_0.0.1.pth, then place it under theComfyUI/models/TTS/DAC.speech.v1.0directory.