https://github.com/vekitan55/SimpleFlux1Merger
Simplified and Expert version of Flux.1 merger node for ComfyUI.
https://github.com/vekitan55/SimpleFlux1Merger
Last synced: about 1 month ago
JSON representation
Simplified and Expert version of Flux.1 merger node for ComfyUI.
- Host: GitHub
- URL: https://github.com/vekitan55/SimpleFlux1Merger
- Owner: vekitan55
- License: mit
- Created: 2025-04-23T11:36:49.000Z (about 2 months ago)
- Default Branch: main
- Last Pushed: 2025-04-23T12:09:47.000Z (about 2 months ago)
- Last Synced: 2025-04-23T12:32:49.257Z (about 2 months ago)
- Language: Python
- Homepage:
- Size: 0 Bytes
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
- awesome-comfyui - **Simple Flux.1 Merger for ComfyUI** - based models with intuitive control. This extension provides both simplified group merging and expert per-layer control, including support for advanced difference-based merge modes. Basically the code was generated by ChatGPT. (All Workflows Sorted by GitHub Stars)
README
# Simple Flux.1 Merger for ComfyUI
A custom ComfyUI node set for merging Flux.1-based models with intuitive control.
This extension provides **both simplified group merging** and **expert per-layer control**, including support for advanced difference-based merge modes.
Basically the code was generated by ChatGPT.
---
## 📦 Included Nodes
### 1. **SimplifiedFlux1Merge**
- Intuitive UI with 9 group-based sliders for merging double and single transformer blocks.
- Preserves control over top-level layers like `img_in`, `txt_in`, `final_layer`.
- Designed for quick and effective model blending.
### 2. **ExpertFlux1Merge**
- Fine-grained control for all 19 `double_blocks`, 38 `single_blocks`, and all top-level blocks (`img_in`, `txt_in`, etc.).
- Ideal for advanced users seeking precise model tuning.
- Includes interpolation-safe FP16/BF16 handling and memory optimization.
---
## 🧠 Supported Merge Modes
| Mode | Description |
|-------------------|-----------------------------------------------------------------------------|
| `standard` | Linear interpolation between A and B (`A * w + B * (1-w)`) |
| `add_difference` | Adds difference between B and A to A (`A + w * (B - A)`) |
| `train_difference`| Adds training diff from C to A (`A + w * (B - C)`), mimicking LoRA learning |
---
## 🛠 Installation
Clone or download this repository into your ComfyUI `custom_nodes/` directory:
git clone https://github.com/yourname/comfyui-simplified-flux1.git
---
## 🚀 Usage
Launch ComfyUI.
Add either:
Simplified Flux.1 Merge node
Expert Flux.1 Merge node
Connect 3 Flux.1-compatible models to model_a, model_b, and model_c.
Adjust sliders to define merge ratios.
Select merge mode and execute.
---
## 📌 Notes
Compatible with Flux.1 and Schnell variants (if matching architecture).
Requires sufficient GPU memory; optimized for BF16.
train_difference mode requires model C to match A/B in structure.
---
## 📜 License
MIT License
---
## 🙏 Credits
Based on Flux.1 architecture by Black Forest Labs.
Inspired by [SuperMerger](https://github.com/hako-mikan/sd-webui-supermerger).
Developed for ComfyUI advanced model merging workflows.