ComfyUI Pixtral Large Extension

A ComfyUI custom node that integrates Mistral AI's Pixtral Large vision model, enabling powerful multimodal AI capabilities within ComfyUI. Pixtral Large is a 124B parameter model (123B decoder + 1B vision encoder) that can analyze up to 30 high-resolution images simultaneously.

Features

• 🖼️ Process up to 30 high-resolution images in a single request

• 🧠 Leverages Pixtral Large's 124B parameter architecture

• 📝 Generate detailed descriptions and analysis of images

• 📊 Support for documents, charts, and natural images

• 🌐 128K context window for extensive image processing

• 🔤 Multilingual capabilities including:

  • English
  • Hebrew (עברית)
  • Arabic (العربية)
  • Chinese (中文)
  • Japanese (日本語)
  • Korean (한국어)
  • And many more languages

• 📚 Advanced OCR in multiple languages and scripts

• 🛠️ Customizable parameters for fine-tuned responses

Installation

1. Clone this repository into your ComfyUI's custom_nodes directory:

cd ComfyUI/custom_nodes
https://github.com/ShmuelRonen/ComfyUI_pixtral_large.git

2. Restart ComfyUI

Included Nodes

The extension adds three powerful nodes to ComfyUI:

1. Pixtral Large

Main node for image analysis using Pixtral Large.

Parameters:

• prompt: Your query about the image(s) - can be in any supported language
• images: Input images to analyze
• api_key: Your Mistral AI API key
• temperature: Response randomness (0.0-1.5)
• maximum_tokens: Max response length (1-32768)
• top_p: Nucleus sampling parameter (0.0-1.0)

Use Cases:

• Image analysis and description
• Document text extraction
• Chart and graph interpretation
• Mathematical reasoning
• Cross-lingual image understanding

2. Multi Images Input

Specialized node for combining multiple images into a batch for analysis.

Parameters:

• inputcount: Number of image inputs (2-30)
• Dynamic image inputs (generated based on inputcount)
• Optional parameters for each image slot

Features:

• Automatic batch creation
• Support for up to 30 simultaneous images
• Compatible with all ComfyUI image outputs
• Maintains image quality and resolution
• Efficient memory handling

Use Cases:

• Batch document processing
• Multiple page analysis
• Comparative image analysis
• Sequential image storytelling
• Before/after image analysis

3. Preview Text

Advanced text output display node for viewing Pixtral Large results.

Parameters:

• text: Input text to display (automatically connected to Pixtral Large output)
• Dynamic sizing
• Auto-formatting

Features:

• RTL language support
• Unicode text display
• Formatted output
• Multi-paragraph handling
• Supports all languages
• Copy-paste functionality

Use Cases:

• Displaying analysis results
• Debugging outputs
• Text verification
• Intermediate result inspection
• Documentation generation

Node Connections and Workflow Examples

Basic Single Image Analysis

graph LR
    A[Load Image] --> B[Pixtral Large]
    B --> C[Preview Text]

Multi-Image Analysis

graph LR
    A[Load Image 1] --> C[Multi Images Input]
    B[Load Image 2] --> C
    C --> D[Pixtral Large]
    D --> E[Preview Text]

Complex Document Analysis

graph LR
    A[Load Image 1] --> D[Multi Images Input]
    B[Load Image 2] --> D
    C[Load Image 3] --> D
    D --> E[Pixtral Large]
    E --> F[Preview Text]

Multilingual Capabilities

Pixtral Large offers robust multilingual support for both input and output:

Text Recognition (OCR)

• Recognizes text in multiple scripts and languages
• Particularly strong with:
  • Hebrew (עברית) - including modern and historical texts
  • Latin scripts
  • CJK characters (Chinese, Japanese, Korean)
  • Arabic script
  • Cyrillic
  • And more

Analysis and Response

• Can understand prompts in multiple languages
• Provides responses in the same language as the prompt
• Handles mixed-language content effectively
• Accurate translation and transcription capabilities

Example Usage

# Hebrew prompt example
prompt = "תאר את התמונה בעברית"

# Mixed language example
prompt = "Analyze this image and provide the response in Hebrew (עברית)"

API Key Setup

1. Get your API key from Mistral AI Platform
2. Enter the key in the Pixtral Large node's api_key field

Error Handling

Common error messages and solutions:

Multi Images Input Errors

• "At least 2 images are required": Add more images to input slots
• "Exceeded maximum image count": Reduce number of input images to 30 or less
• "Invalid image format": Ensure images are in supported format

Pixtral Large Errors

• "API Error": Verify API key and internet connection
• "Invalid prompt": Check prompt formatting
• "Token limit exceeded": Reduce maximum_tokens parameter

Preview Text Errors

• "Unicode decode error": Check text encoding
• "Display buffer full": Reduce output size

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

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

Acknowledgments

• Thanks to Mistral AI for providing the Pixtral Large model
• Built for the ComfyUI community

Version History

• 1.0.0: Initial release
  • Full node suite implementation
  • Multi-image support
  • Multilingual capabilities including Hebrew
  • Advanced text preview features

Support

If you encounter any issues or have questions:

1. Check the Issues page
2. Create a new issue if needed