M16 by Freyja Pixel 💖

This is a stable two pass, hand-tuned, Illustrious-based workflow, designed to accurately create and enhance 2048x2048 images with a focus on hentai and anime styling with accurate anatomy and body posture! The more hentai you add, the more morphs you may get~ 💖

💥M16A2 is released and it is the recommended workflow! It includes my notes about Wai-Illustrious version 16, a Readme, my legacy notes about Wai-Illustrious v 8 - 15, AddDetailers for face and hands, VRAM cleaners, Power LoRA Loaders, and a noise buffer! I have packaged dozens of example images and additional developmental workflows as examples in the .zip archive (SFW + NSFW + Morphs/wrong gens for your reference)!

1️⃣ M16 base workflow is more straightforward and easier to use, though less refined! If you choose it, MAKE SURE to change your KSamplers to "Euler_Ancestral" and "Normal" for pass one and "Euler_Ancestral" and "simple" for pass 2!

In general, with Wai-Illustrious 16, my recommended detail LoRAs are less needed and might overbake things at default settings! They also tend to add levels of skin detail and realism! See the Jane Doe (Mouse Girl) examples in the show case for the differences!

It uses Wai-Illustrious v16 (https://civitai.com/models/827184?modelVersionId=2514310) for both passes, with hand-tuned settings for both passes!

It uses ComfyUI Impact pack: https://www.runcomfy.com/comfyui-nodes/ComfyUI-Impact-Pack

It uses the Add-detailer / Facedetailer / Ultralytics Impact subpack: https://github.com/ltdrdata/ComfyUI-Impact-Subpack

It uses ComfyUI Easy Use: https://www.runcomfy.com/comfyui-nodes/ComfyUI-Easy-Use

It uses Differential Diffusion: https://www.runcomfy.com/comfyui-nodes/ComfyUI/DifferentialDiffusion

For Addetail models: https://huggingface.co/ashllay/YOLO_Models/tree/e07b01219ff1807e1885015f439d788b038f49bd/bbox

It uses AnimeSharp4x: https://openmodeldb.info/models/4x-AnimeSharp

OR Remacri's Original Upscaler: https://civitai.com/models/147759/remacri

Other good options for upscale pass include:

RealESREGAN4X_6BANIME_DB: https://openmodeldb.info/models/4x-realesrgan-x4plus-anime-6b

RealESREGAN2x (I tested original, not "plus"): https://openmodeldb.info/models/2x-realesrgan-x2plus

I ENCOURAGE the use of these quality LoRAs:

ADD Detail Slider: https://civitai.com/models/1333749/add-detail-slider?modelVersionId=1506032

Illustrious Stabilizer: https://civitai.com/models/971952?modelVersionId=2055853

I've meticulously hand-tuned and tested all of my previous Wai-Illustrious workflows using this exact refiner and LoRA calibration. I am confident to say.... Experiment away!