ComfyUI-VIBE

ComfyUI-VIBE 🎨

<table> <tr> <td width="320" align="center" valign="top"> <a href="https://youtube.com/shorts/6ChWRx5K2CQ" target="_blank"> <img src="examples/I2I-woman-in-a-hat.png" width="300"><br> <strong>▶️ Play Video Demo</strong> </a> </td> <td valign="top"> Implementation of <b>VIBE</b> (Visual Instruction Based Editor) as a custom node for <a href="https://github.com/comfyanonymous/ComfyUI">ComfyUI</a>. Edit images using natural language instructions (e.g., "make it winter", "change the dog to a cat").<br><br> VIBE leverages the efficient <b>Sana1.5-1.6B</b> diffusion model and <b>Qwen3-VL-2B-Instruct</b> for fast, high-quality image manipulation.<br><br> <h3>✨ Features</h3> <ul> <li><b>Instruction-based Editing:</b> No complex prompting required.</li> <li><b>Latent Support:</b> Connect an <code>Empty Latent Image</code> to define output resolution.</li> <li><b>Fast Inference:</b> Powered by Sana1.5's linear attention.</li> <li><b>Local Model Support:</b> Runs entirely on your hardware.</li> </ul> </td> </tr> </table>

⚡ Easy Installation (via ComfyUI Manager)

1. Load this workflow into ComfyUI.
2. Open ComfyUI Manager and click "Install Missing Custom Nodes".
3. Restart ComfyUI after the installation finishes.
4. Once reloaded, locate the VIBE Image Editor node (in Step 2) and click the "Check / Download Model" button. This will automatically download the necessary weights.

---

🖼️ Example Workflow

Drag and drop this image into ComfyUI to load the workflow:

---

🚀 Installation

1. Clone the repository: Navigate to your ComfyUI/custom_nodes folder and run:

   git clone https://github.com/ato-zen/ComfyUI-VIBE
   

2. Install dependencies: Open terminal in the ComfyUI-VIBE folder and run:

   pip install -r requirements.txt
   

---

📂 Model Setup

This node looks for models in: ComfyUI/models/vibe/

1. Create the target folder:

   mkdir -p models/vibe
   cd models/vibe
   

2. Download the weights: Clone from Hugging Face (requires git-lfs):

   git clone https://huggingface.co/iitolstykh/VIBE-Image-Edit
   

Structure should look like:

📂 ComfyUI/
└── 📂 models/
    └── 📂 vibe/
        └── 📂 VIBE-Image-Edit/
            ├── model_index.json    
            ├── 📂 scheduler/
            ├── 📂 text_encoder/
            ├── 📂 tokenizer/
            ├── 📂 transformer/
            └── 📂 vae/

---

📜 Credits & Acknowledgements

• Original Project: VIBE: Visual Instruction Based Editor
• Model Authors: Grigorii Alekseenko, Aleksandr Gordeev, Irina Tolstykh, et al.
• Based on: Sana and Qwen3-VL.

ComfyUI Node implementation by ato-zen.