Python desktop and CLI wrapper for running a ComfyUI SynthID-removal workflow.

This app does not reimplement ComfyUI internals. It manages a ComfyUI subprocess, uploads images through ComfyUI's REST API, queues workflow API JSON, polls completion from /history/{prompt_id}, downloads the output through /view, and displays a before/after comparison.

1. Create and activate a Python environment.

python -m venv venv
venv\Scripts\activate
pip install -r requirements.txt

2. Install ComfyUI.

Automatic:

python download_models.py --install-comfy

Manual:

git clone https://github.com/comfyanonymous/ComfyUI.git ComfyUI
cd ComfyUI
pip install -r requirements.txt
cd ..

If ComfyUI is elsewhere, set COMFYUI_DIR before launching this app.

3. Download models into ComfyUI.

python download_models.py

Expected paths:

ComfyUI/models/diffusion_models/z_image_turbo_bf16.safetensors
ComfyUI/models/text_encoders/qwen_3_4b.safetensors
ComfyUI/models/vae/ae.safetensors
ComfyUI/models/controlnet/Z-Image-Turbo-Fun-Controlnet-Union-2.1-2602-8steps.safetensors
ComfyUI/models/ultralytics/yolov8n-face.pt

4. Install the ComfyUI custom nodes required by your working workflow:

• Z-Image model loader support if your ComfyUI build does not already include it
• ControlNet preprocessor node providing CannyEdgePreprocessor or Canny
• Face/inpaint nodes providing UltralyticsDetectorProvider, BboxDetectorForEach, InpaintCrop, and InpaintStitch

The app validates required node class names through /object_info before queueing.

GUI:

python main.py

CLI:

python main.py --headless --input image.png
python main.py --headless --input ./folder --output ./results
python main.py --check

The generated API workflow follows this structure:

• Load Image -> VAE Encode
• Load CLIP qwen_3_4b.safetensors -> positive/negative CLIP Text Encode
• Load Diffusion Model z_image_turbo_bf16.safetensors
• Load ControlNet -> Canny preprocessing -> Apply ControlNet
• KSampler: Euler/simple, default steps 9, cfg 1.0, denoise 0.2
• VAE Decode -> Save Image
• Optional face pass: YOLO face detector -> bbox detector -> inpaint crop -> low-denoise KSampler -> inpaint stitch -> Save Image

Output format and quality are applied after ComfyUI returns the generated image. "# Sythn"