ComfyUI-Gemma4

ComfyUI custom nodes integrating Gemma 4 multimodal AI — supporting text generation, image understanding, and video understanding.

Nodes

Node	Description
Gemma 4 Multimodal	Unified node supporting text-only, image, and video input
Gemma 4 Text Generation	Text-only generation
Gemma 4 Image Understanding	Single image or image batch (sequential frames) understanding
Gemma 4 Video Understanding	Video frame understanding with frame sampling

Model Setup

Download the model from ModelScope and place it in:

ComfyUI/models/LLM/gemma-4-12b-it/

Directory structure

models
  |- LLM
    |- gemma-4-12b-it
      |- chat_template.jinja
      |- config.json
      |- configuration.json
      |- generation_config.json
      |- model.safetensors
      |- processor_config.json
      |- README.md
      |- tokenizer.json
      |- tokenizer_config.json

Installation

1. Clone this repository into ComfyUI/custom_nodes/
2. Install dependencies: pip install -r requirements.txt
3. Restart ComfyUI

License

Apache-2.0