Ultimate ComfyUI SDXL/PDXL/Illustrious Workflow — v1.2

Use Ultimate ComfyUI SDXL/PDXL/Illustrious Workflow to generate and refine images from text prompts or input images in txt2img or img2img mode, returning upscaled and refined images through three stages with OpenPose ControlNet and detailers.

Use one stylized model for initial generation and a secondary model for refinement and upscaling across three stages.

Detailers cover hands, breasts, penis, butt, vagina, face, eyes/mouth, and jewelry; bbox and SAM provide masks, OpenPose ControlNets guide hand and finger refinements, and MediaPipe FaceMesh detects face features individually or together.

Provide a text prompt or load an image, then choose txt2img or img2img with optional inversion and tinting.

It handles width and height automatically for conditioning and scaling and provides three sample/upscale stages.

To use the detailers, download the required bbox models from Hugging Face or Civitai and get the SAM model through ComfyUI's model manager.

The listed node packs are: comfyroll, comfyui_controlnet_aux, comfyui-custom-scripts, comfyui-dream-project, comfyui-easy-use, comfyui-impact-pack, comfyui-impact-subpack, comfyui-inspire-pack, ComfyUI-TiledDiffusion, and ComfyUI-YOLO.

The listed model files are: 4xNomosWebPhoto_RealPLKSR.pth, AssImplantsIL.safetensors, Betterhand1llust.safetensors, breast20size20slider20.h5T3.safetensors, diffusion_pytorch_model.safetensors, eye-000007.safetensors, illustrij_v21.safetensors, Innie_pussy_v2.safetensors, sam_vit

The bbox files are: bbox/adetailerJewelry_v10.pt, bbox/assdetailer-seg.pt, bbox/face_yolov8m.pt, bbox/female-breast-v4.7.pt, bbox/nipple.pt, bbox/penisV2.pt, bbox/PitHandDetailer-v2-Test-v9c.pt, and bbox/pussyV2.pt.

Use the control panel's dropdowns and boolean toggles to adjust models and settings.

Set scaling factors independently for each stage, and bypass stage 2 or 3 positive prompts, negative prompts, or both to reuse prompts across all stages.

Use separate LoRA stacks for stage 1, stages 2 and 3, and detailers; Pony toggles can add scores to prompts, and the text can be edited or replaced for embeddings.

If you bypass a detailer, its group, or its model loader, keep the get_DETAILER_*_IMAGE nodes enabled.

Suggestion · non vérifié

No Pipe version is specified for this workflow.