ComfyUI Extension: ComfyUI-JM-Gemini-API

Authored by synthetai

Created

Updated

3 stars

A custom node for ComfyUI that generates images using Google's Gemini API, supporting both text-to-image and image-to-image generation.

Custom Nodes (0)

    README

    ComfyUI-JM-Gemini-API

    Custom nodes for ComfyUI that generate images and videos using Google's Gemini API, supporting text-to-image, image-to-image, text-to-video, and image-to-video generation.

    English | 简体中文

    Features

    Image Generation

    • Support for multiple Gemini image models:
      • gemini-3-pro-image-preview (default, with 2K resolution)
      • gemini-2.5-flash-image
    • Text-to-image generation
    • Image-to-image generation (single or multiple input images)
    • Image editing mode (single image input)
    • Configurable aspect ratios (1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9)
    • Resolution control (1K, 2K, 4K) - only for gemini-3-pro-image-preview
    • Support up to 10 input images
    • Automatic image saving to ComfyUI output directory

    Video Generation

    • Support for Gemini Veo video models:
      • veo-3.1-generate-preview (default)
      • veo-3.1-fast-generate-preview
      • veo-3.0-generate-001
      • veo-3.0-fast-generate-001
    • Text-to-video generation
    • Image-to-video generation (animate a single image)
    • First-last frame interpolation (Veo 3.1 only)
    • Negative prompt support
    • Configurable aspect ratios (16:9, 9:16)
    • Resolution control (720p, 1080p)
    • Duration control (4, 6, 8 seconds)
    • Automatic video saving to ComfyUI output directory

    Important Limitations:

    • 1080p resolution: Only supports 8-second duration for Veo 3.1 models
    • First-last frame interpolation: Only available for Veo 3.1 models with 8-second duration

    Installation

    1. Clone or download this repository to your ComfyUI custom_nodes directory:
    cd custom_nodes
git clone https://github.com/synthetai/ComfyUI-JM-Gemini-API.git
    1. Install required dependencies:
    cd ComfyUI-JM-Gemini-API
pip install -r requirements.txt
    1. Restart ComfyUI

    Requirements

    • Python 3.8+
    • ComfyUI
    • google-genai >= 0.2.0
    • Pillow >= 10.0.0
    • torch
    • torchvision
    • numpy
    • httpx[socks] (for proxy support)

    Proxy Configuration (Optional)

    If you need to use a proxy to access the Gemini API (common for users in China), you can set environment variables:

    HTTP/HTTPS Proxy

    export HTTP_PROXY="http://your-proxy:port"
export HTTPS_PROXY="http://your-proxy:port"

    SOCKS5 Proxy

    export HTTP_PROXY="socks5://your-proxy:port"
export HTTPS_PROXY="socks5://your-proxy:port"

    Or set them before starting ComfyUI:

    HTTP_PROXY="socks5://127.0.0.1:1080" HTTPS_PROXY="socks5://127.0.0.1:1080" python main.py

    Usage

    Getting Gemini API Key

    1. Visit Google AI Studio
    2. Sign in with your Google account
    3. Create a new API key
    4. Copy the API key for use in the node

    Node Parameters

    Required Inputs:

    • gemini_api_key: Your Gemini API key (string)
    • prompt: Text prompt describing the image you want to generate (multiline text)
    • model: Choose between:
      • gemini-3-pro-image-preview (default, supports 1K/2K/4K resolution)
      • gemini-2.5-flash-image (faster, aspect ratio only)
    • aspect_ratio: Image aspect ratio (1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9)
    • resolution: Image resolution (1K, 2K, 4K)
      • Note: Only effective for gemini-3-pro-image-preview model
      • Default: 2K

    Optional Inputs:

    • image1 ~ image10: Up to 10 optional image inputs for image-to-image generation
      • Connect output from Load Image node
      • Can be left empty for text-to-image generation

    Outputs:

    • image: Generated image (ComfyUI IMAGE tensor format)
      • Can be connected to Preview Image or Save Image nodes
      • Automatically saved to ComfyUI output directory

    Usage Examples

    1. Text-to-Image Generation

    1. Add "JM Gemini Image Generator" node to your workflow
    2. Enter your Gemini API key
    3. Write your prompt
    4. Select model and aspect ratio
    5. Leave all image inputs empty
    6. Connect output to Preview Image node
    7. Run the workflow

    2. Image Editing (Single Image)

    1. Add Load Image node and load your image
    2. Add "JM Gemini Image Generator" node
    3. Connect Load Image output to image1 input
    4. Enter editing instructions in prompt (e.g., "Add a sunset background")
    5. Configure model and parameters
    6. Run the workflow

    3. Image-to-Image (Multiple Images)

    1. Add multiple Load Image nodes
    2. Connect them to image1, image2, etc.
    3. Enter a prompt describing how to combine/transform the images
    4. Run the workflow

    Model Differences

    gemini-3-pro-image-preview

    • Supports resolution parameter (1K, 2K, 4K)
    • Default resolution: 2K
    • Supports single image editing
    • Higher quality output

    gemini-2.5-flash-image

    • Faster generation
    • Uses aspect ratio only (no resolution parameter)
    • Good for quick iterations

    Troubleshooting

    Common Issues

    1. "Gemini API key is required"

      • Make sure you've entered a valid API key

    2. "No images were generated"

      • Check your prompt is clear and descriptive
      • Try a different aspect ratio or resolution
      • Verify your API key is valid and has sufficient quota

    3. "Failed to generate image"

      • Check your internet connection
      • Verify API key permissions
      • Check Gemini API service status

    4. Image quality issues

      • For gemini-3-pro-image-preview, try using different resolutions (1K, 2K, or 4K)
      • Make your prompt more detailed and specific

    Output Directory

    Generated images are automatically saved to:

    • ComfyUI/output/ directory
    • Filename format: {model}_{mode}_{timestamp}.png
      • Example: gemini3pro_text2img_1234567890.png

    License

    MIT License

    Credits

    Developed by JM

    Based on Google's Gemini API

    Support

    For issues and feature requests, please visit the GitHub repository

    Video Node Usage

    Node: JM Gemini Video Generator

    Required Inputs:

    • gemini_api_key: Your Gemini API key (string, encrypted input)
    • prompt: Text prompt describing the video you want to generate (multiline text)

    Optional Inputs:

    • negative_prompt: Describe what you don't want in the video (multiline text)
    • model: Choose video model:
      • veo-3.1-generate-preview (default, highest quality)
      • veo-3.1-fast-generate-preview (faster generation)
      • veo-3.0-generate-001 (stable version)
      • veo-3.0-fast-generate-001 (stable + fast)
    • aspect_ratio: Video aspect ratio (16:9 or 9:16, default: 16:9)
    • resolution: Video resolution (720p or 1080p, default: 720p)
    • duration: Video duration in seconds (4, 6, or 8, default: 8)
    • first_image: First frame image (optional, for image-to-video or interpolation)
    • last_image: Last frame image (optional, only for Veo 3.1 interpolation)

    Output:

    • video_path: Path to the generated video file (STRING)

    Video Generation Modes

    1. Text-to-Video

    1. Add "JM Gemini Video Generator" node
    2. Enter API key and prompt
    3. Leave both image inputs empty
    4. Configure model and parameters
    5. Run workflow - video will be saved to output directory

    2. Image-to-Video

    1. Add Load Image node and load your image
    2. Connect it to first_image input
    3. Enter prompt describing the motion/animation
    4. Run workflow

    3. First-Last Frame Interpolation (Veo 3.1 only)

    1. Add two Load Image nodes
    2. Connect first image to first_image input
    3. Connect last image to last_image input
    4. Select a Veo 3.1 model (veo-3.1-generate-preview or veo-3.1-fast-generate-preview)
    5. Set duration to 8 seconds (required for interpolation mode)
    6. Enter prompt describing the transition
    7. Run workflow - the model will generate smooth interpolation between frames

    Important Notes:

    • Video generation can take several minutes. The node will poll the API every 10 seconds with a 20-minute timeout.
    • 1080p resolution is only supported with 8-second duration for Veo 3.1 models
    • First-last frame interpolation requires Veo 3.1 models and 8-second duration only

    Changelog

    Version 1.1.0

    • Added video generation support with JM Gemini Video Generator node
    • Support for Veo 3.1 and Veo 3.0 models
    • Text-to-video generation
    • Image-to-video animation
    • First-last frame interpolation (Veo 3.1 only)
    • Refactored code structure with separate node files and shared utilities

    Version 1.0.0

    • Initial release
    • Support for gemini-3-pro-image-preview and gemini-2.5-flash-image
    • Text-to-image generation
    • Image-to-image generation (up to 10 images)
    • Configurable aspect ratios and resolutions