This repository provides a comprehensive suite of custom nodes for ComfyUI, aimed at extending its capabilities for AI-powered creative workflows. It targets users looking to integrate advanced features like screen sharing, real-time video manipulation, GPT-powered interactions, speech recognition, text-to-speech, and 3D model processing directly within their ComfyUI setups. The primary benefit is the significant expansion of ComfyUI's functionality into more interactive and application-like use cases.

How It Works

The nodes leverage ComfyUI's extensible architecture to introduce new input/output types and processing capabilities. Key features include a "Workflow-to-APP" system that transforms ComfyUI workflows into standalone web applications, real-time screen capture and video streaming nodes, and deep integration with various Large Language Models (LLMs) for text generation and prompt manipulation. It also incorporates advanced image processing, 3D model handling, and speech technologies, offering a unified platform for multimodal AI applications.

Quick Start & Requirements

• Install: Clone the repository into your ComfyUI custom_nodes directory:

  cd ComfyUI/custom_nodes
  git clone https://github.com/shadowcz007/comfyui-mixlab-nodes.git
  

• Requirements: Install dependencies using install.bat (Windows) or pip install -r requirements.txt (ensure your Python environment is activated).
• Prerequisites: ComfyUI, Python 3.11, PyTorch 2.3.1+cu121. Some nodes may require additional model downloads (e.g., for speech, LLMs, inpainting).
• Docs: Mixlab Community

Highlighted Details

• Workflow-to-APP: Convert ComfyUI workflows into shareable web applications with custom UIs.
• Real-time Interaction: ScreenShareNode and FloatingVideoNode enable real-time pixel stream capture and display.
• LLM Integration: Supports multiple LLMs (local and cloud-based) for text generation, prompt completion, and agent-like behavior.
• Multimodal Capabilities: Integrates speech recognition, TTS, 3D model loading, and advanced image manipulation.

Maintenance & Community

• Community: Chinese community access via www.mixcomfy.com and WeChat public account "Mixlab 无界社区".
• Contact: Business cooperation via 389570357@qq.com.

Licensing & Compatibility

• License: Not explicitly stated in the README.
• Compatibility: Designed for ComfyUI. Some model usage may have separate licensing restrictions (e.g., BRlA AI's briarmbg for non-commercial use).

Limitations & Caveats

The project mentions potential Pillow conflicts with simple-lama-inpainting and advises using HTTPS for network services. Some advanced features like dynamic delay based on text length are noted as additions. The "Workflow-to-APP" feature requires the plugin directory to be named comfyui-mixlab-nodes precisely.