CumfyUI Workflow Enhancements

This repository hosts a tuned ComfyUI workspace optimized for explicit/mature image generation workflows. It includes ready-to-run nodes, scripts, and reference workflows focused on IPAdapter, ControlNet, and CASE-prompt tooling.

Contents

• custom_nodes/ComfyUI_IPAdapter_plus/

  • Upstream IPAdapter+ extension for style/subject transfer.
  • Additional SDXL models downloaded into ComfyUI/models/ipadapter/:
    • ip-adapter-plus_sdxl_vit-h.safetensors
    • ip-adapter-plus-face_sdxl_vit-h.safetensors
    • ip_plus_composition_sdxl.safetensors

• custom_nodes/ComfyUI-Manager/, ComfyUI-Lora-Manager/, ComfyUI-EasyCivitai-XTNodes/

  • Utilities for model management within ComfyUI.

• models/ (ignored by Git)

  • Large assets, including checkpoints, CLIP vision encoders, ControlNet weights.
  • Currently populated with:
    • models/clip_vision/CLIP-ViT-H-14-laion2B-s32B-b79K.safetensors
    • models/clip_vision/CLIP-ViT-bigG-14-laion2B-39B-b160k.safetensors
    • models/ipadapter weights listed above.
    • models/controlnet/OpenPoseXL2.safetensors (SDXL openpose ControlNet)

• workflows/

  • Ready-to-run JSON workflows and recipe scripts.
  • Key entries:
    • illustrious_case_template.json – base workflow anchored on Pony Diffusion V6 XL + ExpressiveH.
    • illustrious_case_ipadapter.json, illustrious_case_ipadapter_controlnet.json, my_explicit_scene.json – generated variants demonstrating IPAdapter & ControlNet integrations.
    • illustrious_case_romantic.json – romantic preset without ControlNet for softer compositions.
    • illustrious_case_aftercare.json – aftercare portrait using the IPAdapter Plus Face preset.
    • illustrious_case_img2img.json – image-to-image variant: load a source image via input/source_image.png, preserves pose with ControlNet and style with IPAdapter.
    • illustrious_case_img2img_enhanced.json – two-pass img2img with latent upscaling and a second sampler (DPM++ 2M Karras) for higher fidelity.
    • recipes/case_prompt_pools.json – curated CASE vocab and style presets.
    • recipes/generate_case_prompt.py – CLI helper for building CASE prompts (--style-preset romantic|hardcore|aftercare).
    • recipes/apply_case_prompt.py – workflow patcher with flags:
      • --style-preset, --mature, --neg-groups ...
      • --add-ipadapter, --ipadapter-image, --ipadapter-weight, --ipadapter-community
      • --add-controlnet, --controlnet, --control-image, strength/start/end windows
      • LoRA/CLIP tuning (--lora-strength-model, --lora-strength-clip, --clip-layer)

• workflows/CASE_prompt_notes.md

  • Documentation on CASE prompting, style presets, LoRA and ControlNet usage.
  • Highlights command examples and locations for necessary assets.

• scripts/

  • Helper shell scripts from upstream ComfyUI for downloading or verifying models.

• user/

  • Per-ComfyUI settings/database (included for completeness).

Step-by-Step Guide

Below is the exact sequence used to build this workspace from a vanilla ComfyUI clone. Run commands from the repository root unless noted.

1. Install IPAdapter+ and supporting managers

   git clone https://github.com/cubiq/ComfyUI_IPAdapter_plus custom_nodes/ComfyUI_IPAdapter_plus
   git clone https://github.com/ltdrdata/ComfyUI-Manager custom_nodes/ComfyUI-Manager
   git clone https://github.com/willmiao/ComfyUI-Lora-Manager custom_nodes/ComfyUI-Lora-Manager
   git clone https://github.com/kohya-ss/ComfyUI-EasyCivitai-XTNodes custom_nodes/ComfyUI-EasyCivitai-XTNodes

   Remove nested .git folders so the node suites live directly inside this repository.

2. Download large model assets (kept out of Git)

   • Optional pose libraries can be unzipped into ComfyUI/input/pose_library/ (e.g. posesPacksCollection_portraitV2); the helper scripts will accept them via --control-image.

   # SDXL IPAdapter weights
   curl -L -o models/ipadapter/ip-adapter-plus_sdxl_vit-h.safetensors \
        https://huggingface.co/h94/IP-Adapter/resolve/main/sdxl_models/ip-adapter-plus_sdxl_vit-h.safetensors
   curl -L -o models/ipadapter/ip-adapter-plus-face_sdxl_vit-h.safetensors \
        https://huggingface.co/h94/IP-Adapter/resolve/main/sdxl_models/ip-adapter-plus-face_sdxl_vit-h.safetensors
   curl -L -o models/ipadapter/ip_plus_composition_sdxl.safetensors \
        https://huggingface.co/ostris/ip-composition-adapter/resolve/main/ip_plus_composition_sdxl.safetensors
   
   # CLIP vision encoders for SDXL IPAdapter
   curl -L -o models/clip_vision/CLIP-ViT-H-14-laion2B-s32B-b79K.safetensors \
        https://huggingface.co/h94/IP-Adapter/resolve/main/models/image_encoder/model.safetensors
   curl -L -o models/clip_vision/CLIP-ViT-bigG-14-laion2B-39B-b160k.safetensors \
        https://huggingface.co/h94/IP-Adapter/resolve/main/sdxl_models/image_encoder/model.safetensors
   
   # SDXL OpenPose ControlNet
   curl -L -o models/controlnet/OpenPoseXL2.safetensors \
        https://huggingface.co/thibaud/controlnet-openpose-sdxl-1.0/resolve/main/OpenPoseXL2.safetensors

   Place checkpoints (Pony Diffusion V6 XL, ExpressiveH LoRA, etc.) inside the existing models/ subdirectories.

3. Update CASE prompt tooling – edit the files under workflows/recipes/:

   • case_prompt_pools.json: add mature vocabulary and style presets.
   • generate_case_prompt.py: expose --style-preset and bias sampling toward the chosen modes.
   • apply_case_prompt.py: add LoRA/CLIP tuning, ControlNet, and IPAdapter automation flags.

4. Create tailored workflows – start from workflows/illustrious_case_template.json and run:

   workflows/recipes/apply_case_prompt.py \
     --workflow workflows/illustrious_case_template.json \
     --output workflows/my_explicit_scene.json \
     --style-preset hardcore --mature \
     --neg-groups body_artifacts fluid_control \
     --add-ipadapter --ipadapter-image ComfyUI/output/openwebui_00001_.png \
     --ipadapter-weight 0.9 --ipadapter-end 0.65 \
     --add-controlnet --controlnet OpenPoseXL2.safetensors \
     --control-image ComfyUI/input/control_reference.png

   The helper copies referenced images into ComfyUI/input/ and rewires the sampler/conditioners.

5. Document and commit – update .gitignore, add README notes, then commit & push to GitHub.

Following these steps reproduces the configuration committed here from scratch.

Recommended CivitAI References

• Prompt Notebook
• ComfyUI Workflow: Study Case
• Scene Composer: Overview & Usecases
• OpenPose NSFW Pose Pack (Portrait V2) – portrait pose library unzipped into ComfyUI/input/pose_library/.

Explore the CivitAI workflows tab for additional explicit-focused templates (search terms like "Illustrious", "Pony", or "NSFW ComfyUI").

Quick Start

1. Install dependencies (Python, ComfyUI runtime) per upstream ComfyUI instructions or use the provided scripts.
2. Populate ComfyUI/models/ with checkpoints/LoRAs as needed (see .gitignore for directories expected to remain local).
3. Generate a workflow tailored to your references:

   workflows/recipes/apply_case_prompt.py \
     --workflow workflows/illustrious_case_template.json \
     --output workflows/my_scene.json \
     --style-preset hardcore --mature \
     --neg-groups body_artifacts fluid_control \
     --add-ipadapter --ipadapter-image /path/to/style.png \
     --ipadapter-weight 0.9 --ipadapter-end 0.65 \
     --add-controlnet --controlnet OpenPoseXL2.safetensors \
     --control-image /path/to/pose.png

4. Import the resulting JSON in ComfyUI and tweak LoRA/IPAdapter/ControlNet weights via the node UI.

Notes

• Large assets (*.safetensors, checkpoints, embeddings, etc.) are ignored from Git and must be downloaded manually with the provided scripts or commands.
• Macro-level gitignore excludes output/input directories to keep repo size manageable.
• The commit history includes the full ComfyUI node suites cloned from upstream repositories.
• Ensure you have appropriate hardware (GPU with sufficient VRAM) for SDXL checkpoints and IPAdapter operations.