comfy-gemini

ComfyUI Gemini Nodes

This repository provides custom nodes for ComfyUI that leverage the Google Gemini API for prompt generation and enhancement.

Installation

1. Clone this repository into your ComfyUI/custom_nodes/ directory:

   cd ComfyUI/custom_nodes/
   git clone https://github.com/your-username/comfy-gemini.git # Replace with the actual repo URL if different
   

2. Install the required dependencies:

   cd comfy-gemini
   pip install -r requirements.txt
   

3. Restart ComfyUI.

Nodes

1. Gemini Image To Prompt (Gemini/Image)

This node analyzes an input image using a Gemini model and generates a descriptive text prompt based on its content and a provided system prompt template.

Inputs:

• image (IMAGE): The input image to analyze.
• model (STRING): The Gemini vision model to use (e.g., gemini-2.0-pro-latest).
• api_key (STRING): Your Google AI API key.
• prompt_template (STRING, Optional): The system prompt guiding the model on how to describe the image. Defaults to a detailed template focused on illustration regeneration.

Outputs:

• prompt (STRING): The generated text prompt describing the image.

Example Usage: Connect an image output (e.g., from a LoadImage node) to the image input. Provide your API key and select a model. The node will output a text prompt suitable for use in text-to-image generation nodes.

2. Gemini Enhance Prompt (Gemini/Text)

This node takes an existing text prompt and enhances it using a Gemini text model, making it more detailed and descriptive based on a system prompt.

Inputs:

• prompt (STRING): The initial text prompt to enhance.
• api_key (STRING): Your Google AI API key.
• model (STRING): The Gemini text model to use (e.g., gemini-2.5-pro-latest).
• system_prompt (STRING, Optional): The system prompt guiding the model on how to enhance the input prompt. Defaults to a template focused on adding scene, subject, style, and technical details.

Outputs:

• enhanced_prompt (STRING): The enhanced, more detailed text prompt.

Example Usage: Connect a text prompt (e.g., from a Primitive node or another generation node) to the prompt input. Provide your API key and select a model. The node will output an enhanced prompt, potentially leading to more detailed image generations.

Configuration

• API Key: You can provide the Gemini API key directly in the node inputs or set the GEMINI_API_KEY environment variable. Using the environment variable is generally more secure.

Contributing

Contributions are welcome! Please feel free to submit pull requests or open issues.

License

[Specify License, e.g., MIT License]