This is a quantized version of WAI-illustrious-SDXL, Built-in VAE and CLIP included for efficient inference.
All models—Unet, Clip, and Vae—are included in a single ZIP file, complete with sample workflows and a brief guide.
ComfyUI Recommended
Base Model: https://civarchive.com/models/827184/wai-illustrious-sdxl
All credits to the original creator: https://civarchive.com/user/WAI0731
Support & Custom Requests:
Processing, testing, and optimizing these models require significant computational resources and time. If you require a specific Checkpoint or LoRA to be optimized for your hardware constraints, you may submit a formal request and support this ongoing work through the link below:
https://ko-fi.com/morikomorizz
Description
Unet Using Q8_0 For Quant
Clip and Vae Remains at Full precision f16 for Encoder quality
FAQ
Comments (3)
I never though that Illustrious models can do GGUF... What's the capabilities of this? Is this able to run off of a potato of a computer?
on a 4th gen i5 cpu computer, for a 512x512 image it takes approximately 15-20 minutes per image, well at least it works
SDXL, IL/NAI, SD1.5, etc. can be converted to GGUF.
The speed for q8 will not change.
But this allows saving VRAM and hard disk space.
Also, the model loading speed will increase.