A comprehensive ComfyUI workflow for generating high-quality anime-style characters with surgical precision on facial features. This workflow combines the power of SDXL base model with specialized detailers for eyes, face, and lips, plus advanced upscaling capabilities.
✨ KEY FEATURES:
🔹 Multi-Stage Facial Enhancement:
Face Detailer: Uses bbox/adetailers_facef.pt for overall face refinement
Eye Detailer: Specialized bbox/Eyes.pt detector for hyper-detailed irises, reflections, and anatomical accuracy
Lips Detailer: bbox/lips_v1.pt for perfect lip texture, moisture, and shape definition
Each detailer has custom positive/negative prompts optimized for its specific task
🔹 Production-Ready Quality Pipeline:
4x RealESRGAN Anime Upscaling - realesrganX4plusAnime_v1.pt
A/B Comparison Tools - Built-in image comparer nodes for quality assessment
Multi-Stage Saving - Save outputs at different processing stages
Resolution Support - Handles 832x1216 base resolution with upscaling to 1472x2177
🔹 Advanced Prompt Engineering:
Combined prompt system with dual conditioning
Optimized negative prompts for each facial feature
Style-specific prompts for Illustrious aesthetic
⚙️ TECHNICAL REQUIREMENTS:
Required Models:
Required ComfyUI Extensions:
rgthree-comfy (for Power Lora Loader and Image Comparer)
ComfyUI-Impact-Pack (for FaceDetailer and Ultralytics detectors)
ComfyUI_UltimateSDUpscale (for advanced upscaling)
Note: This workflow is heavily optimized for the Illustrious style but can be adapted for other aesthetics by swapping the base checkpoint and adjusting prompts accordingly.
Preview Images: Include 2-3 comparison shots showing:
Base generation without detailers
After face detailing
Final upscaled result
This description is ready for Civitai upload! Want me to adjust anything specific - add more technical details, emphasize certain features, or modify the tone? 😊
