ComfyUI-CSM-Nodes

Custom nodes for ComfyUI implementing the CSM model for text-to-speech generation.

Features

• Node Load CSM Checkpoint: Loads a model checkpoint from ComfyUI/models/sesame/.
• Node Load CSM Tokenizer: Loads a tokenizer from ComfyUI/models/sesame_tokenizer/.
• Node CSM Text-to-Speech: Generates audio from text using the CSM-1B model.
• Node CSM Text-to-Speech with Context: Generates audio from text and example of voice using the CSM-1B model.

Requirements

• Installed ComfyUI.
• Python 3.10+.
• CUDA-compatible GPU (recommended).
• Model weights (ckpt.pt) from sesame/csm-1b in ComfyUI/models/sesame/.
• Tokenizer files (e.g., from Llama-3.2-1B) in ComfyUI/models/sesame_tokenizer/<tokenizer_dir>/.
• Dependencies listed in requirements.txt.

Installation

1. Clone the repository: Clone this repo into custom_nodes folder

   git clone https://github.com/thezveroboy/ComfyUI-CSM-Nodes.git
   

   Install dependencies

   pip install -r requirements.txt
   

   or if you use the portable install, run this in ComfyUI_windows_portable folder

   python_embeded\python.exe -m pip install -r ComfyUI\custom_nodes\ComfyUI-CSM-Nodes\requirements.txt