Custom node for ComfyUI that loads pose reference images from a local folder organized by categories (subfolders). Works great with pose reference packs for AI image generation.
# 1. Go to ComfyUI custom_nodes folder
cd ComfyUI/custom_nodes/
# 2. Clone the repository
git clone https://github.com/simsim9-stack/insane_poses.git
# 3. Restart ComfyUIOr just copy the insane_poses/ folder into ComfyUI/custom_nodes/.
By default the node looks for poses in ../posemy_downloader/poses/.
To change the poses location, set the environment variable:
set INSANE_POSES_DIR=C:\path\to\your\posesOr edit DEFAULT_POSES_DIR in pose_loader.py.
- Category:
Pose - Inputs:
category— dropdown from subfolder namespose— filename (without extension) or full file pathposes_dir— optional override for poses folder
- Outputs:
IMAGE— loaded pose image (tensor)pose_name— filename stem (string)width,height— image dimensions
- Shows pose count and a list of pose names in a category
The repository includes a batch processing script:
python comfyui_batch_processor.pyIt runs every PNG in the poses folder through a ComfyUI workflow (Flux2 + Mannequin LoRA) and saves the output as JPG next to the original.
start_pose_batch.batSupporting modes:
- Full run (all poses)
- Resume from N-th pose
- Test mode (1 pose only)
posemy_downloader/
├── poses/ # Pose images
│ ├── fighting-pose-reference/ # Category = subfolder
│ │ ├── 1.png
│ │ └── ...
│ ├── anime-action-poses/
│ └── ...
├── comfyui_batch_processor.py # Batch processing script
└── start_pose_batch.bat # Batch launcher
insane_poses/ # ComfyUI custom node
├── __init__.py # Node registration
├── pose_loader.py # Node classes
└── README.md # This file
Batch processing uses the workflow in posemy_downloader/1.json:
- Flux2 + ReferenceLatent (Positive → ReferenceLatent, ZeroOut → ReferenceLatent)
- Mannequin LoRA (klein/Mannequin_V1_F29B.safetensors)
- Model: flux-2-klein-9b-fp8.safetensors
- CLIP: qwen_3_8b_fp8mixed.safetensors
- VAE: full_encoder_small_decoder.safetensors
MIT