Scene Composer for ComfyUI

An intuitive, all-in-one node for ComfyUI that brings a powerful, layer-based regional prompting workflow directly into your graph. Say goodbye to managing countless Conditioning (Set Area) nodes and hello to drawing your creative vision.

Created by 834t.

---

✨ Features

The Scene Composer is more than just a node; it's a complete toolkit designed to streamline your workflow and unlock new creative possibilities.

• 🎨 Layer-Based Drawing Interface: Work with multiple layers, just like in your favorite image editor. Each layer is represented by a color.
• ✍️ Per-Layer Control: Assign a unique prompt, opacity, and visibility to each layer for precise control over your composition.
• 🛠️ Intuitive Tools: Comes equipped with a Brush, Line, and Eraser tool.
• 🧠 Smart Mask Normalization: An optional (normalize_masks) feature that intelligently blends overlapping regions, preventing concept "burn-in" and creating more coherent images.
• 📝 Automatic Base Prompt: Optionally (add_base_prompt) create a global conditioning from all layer prompts, helping to unify the overall style and theme of the image.
• 💾 Save & Load Scene State:
  • Save your entire scene (all drawings, prompts, and settings) to a .json file with the "S" button.
  • Load a scene from a file with the "L" button.
  • Your scene is automatically saved within the ComfyUI workflow itself, so you'll never lose your work!

---

⚙️ Installation

1. Navigate to your ComfyUI/custom_nodes/ directory.
2. Clone this repository:

   git clone https://github.com/834t/ComfyUI_834t_scene_composer.git
   

3. Restart ComfyUI.

---

🚀 How to Use

1. Add the Node: Right-click on the canvas and select Add Node > 834t_Nodes > Scene Composer (834t).
2. Set Dimensions: Adjust the width and height widgets to your desired output size.
3. Select a Layer: Click on a color in the palette to activate a layer.
4. Write a Prompt: In the text area, write the prompt for the selected layer (e.g., "a beautiful mountain range").
5. Draw: Use the brush, line, or eraser tools to draw the mask for that concept on the canvas.
6. Adjust Opacity: Use the opacity slider to control the influence of the layer.
7. Repeat: Select other colors to create new layers for other concepts (e.g., "a serene lake in the foreground", "a sky full of stars").
8. Connect:
   • Connect the latent output to your KSampler's latent_image input.
   • Connect the positive output to your KSampler's positive input.
9. Queue Prompt and watch the magic happen!

---

🔌 Node Inputs & Outputs

| Type | Name | Description | | :---- | :---------------- | :------------------------------------------------------------------------------------------------------ | | IN | clip | The CLIP model for text encoding. | | IN | width/height | The dimensions of the output latent space and drawing canvas. | | IN | normalize_masks | (Boolean) If true, normalizes the weight of overlapping masks to prevent artifacts. Recommended: True. | | IN | add_base_prompt | (Boolean) If true, creates a global conditioning from all prompts to improve coherence. | | IN | scene_data | Internal data managed by the widget. You don't need to connect anything here. | | OUT | latent | An empty latent image of the specified dimensions. | | OUT | positive | The combined positive conditioning from all your drawn layers. |

---

🙏 Acknowledgements

This node was created by 834t.

A special and heartfelt thank you goes to my AI assistant and consultant, Gemini 2.5 Pro. thank you Gemini for help in brainstorming features, assist in coding and preparing the documentation. It was a true partner in development.

---

📜 License

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