ComfyUI_Imagen

A custom node for ComfyUI that leverages the Google Cloud Vertex AI Imagen API to generate and edit images.

Installation

1. Clone this repository into your custom_nodes folder.

   cd ComfyUI/custom_nodes
   git clone https://github.com/ru4ls/ComfyUI_Imagen.git

2. Install the required dependencies:

   pip install -r ComfyUI_Imagen/requirements.txt

Project Setup

To use this node, you need a Google Cloud Project with the Vertex AI API enabled.

1. Enable the Vertex AI API: Follow the instructions in the Google Cloud documentation to enable the API for your project.

2. Authenticate Your Environment: This node uses Application Default Credentials (ADC) to securely authenticate with Google Cloud. Run the following gcloud command in your terminal to log in and set up your credentials. This is a one-time setup.

   gcloud auth application-default login

   The node authenticates directly through the installed Python libraries and does not depend on the gcloud.cmd executable being available in your system's PATH at runtime.

3. Create a .env file: In the ComfyUI_Imagen/config/ directory, create a .env file by copying the config/.env.example file. Replace the placeholder values with your Google Cloud project ID and desired location.

   # ComfyUI_Imagen/config/.env
   
   PROJECT_ID="YOUR_PROJECT_ID_HERE"
   LOCATION="YOUR_LOCATION_HERE"
   

Nodes

This package provides a single unified node, Google Imagen, which can perform both text-to-image generation and image editing depending on the inputs provided.

Inputs

• prompt (STRING): The text prompt for image generation or manipulation.
• model_name (STRING): The generation model to use. The list is populated from data/models.json.
• image (IMAGE, optional): An optional input for a base image. Connecting an image switches the node to image-editing mode.
• mask (MASK, optional): An optional mask for inpainting or outpainting. If no mask is provided in editing mode, the node will perform a mask-free, prompt-based image edit.
• aspect_ratio (STRING, optional): The desired aspect ratio for text-to-image generation.
• edit_mode (STRING, optional): The editing mode to use when an image is provided. Options are inpainting and outpainting.
• seed (INT, optional): A seed for reproducibility. See the Seed and Watermarking section below for important usage details.
• add_watermark (BOOLEAN, optional): A toggle to control watermarking and seed behavior.

Outputs

• image (IMAGE): The generated or edited image.

Features and Limitations

Seed and Watermarking

The Google Imagen API does not support using a seed when the automatic watermark feature is enabled.

• For Text-to-Image: To use a seed, you must set the add_watermark toggle to False. When add_watermark is True (the default), the seed value is ignored.
• For Image Editing: The API does not support disabling watermarks. However, to maintain consistent behavior, the add_watermark toggle still controls the seed. Set it to False to use a seed for your edits.

Image Editing

The node supports both masked and mask-free image editing.

• With a Mask: Provide an image and a mask to perform standard inpainting or outpainting.
• Without a Mask: Provide only an image and a prompt. The node will perform a mask-free, prompt-guided edit of the entire image.

Example Usage

Text to Image Generation

1. Add the Google Imagen node to your workflow.
2. Enter a prompt.
3. Ensure no image input is connected.
4. To use a seed, set add_watermark to False.
5. Connect the output image to a PreviewImage or SaveImage node.

Inpainting

1. Add the Google Imagen node.
2. Connect a LoadImage node to the image input.
3. Optional Create a mask and connect it to the mask input.
4. Enter a prompt describing the desired changes.
5. Set edit_mode to inpainting.
6. Connect the output image to a PreviewImage or SaveImage node.

Outpainting

This workflow is identical to inpainting, but with edit_mode set to outpainting and a mask that defines the area to be extended.

Changelog

2025-09-14: Refactored Authentication and Node Logic

• New Authentication: Replaced insecure gcloud subprocess calls with the official google-auth Python library, using Application Default Credentials (ADC). This improves security and removes the dependency on the gcloud executable being in the system's PATH.
• Unified Node: Consolidated the workflow into a single, powerful Google Imagen node that intelligently switches between text-to-image and image-editing modes based on whether an image is provided.
• Watermark & Seed Control: Added an add_watermark toggle. Disabling this allows the seed parameter to be used for reproducible results.
• Enhanced Image Editing: The node now supports both masked (inpainting/outpainting) and mask-free, prompt-guided image editing.
• Centralized Configuration: The .env file is now loaded from the /config directory, and credential loading is handled centrally for a cleaner and more robust startup.
• Improved Error Handling: Implemented error handling that prevents the ComfyUI workflow from crashing on API or configuration errors, instead printing clear messages to the console.

License

This project is licensed under the MIT License - see the LICENSE file for details.