>In this video, I dive into the exciting new features of SDXL 1, the latest version of the Stable Diffusion XL:
High-Resolution Training: SDXL 1 has been trained with high-resolution images to provide outstanding outputs.
Combined Checkpoints: It operates with a combination of two checkpoints – a base model having 3.3 billion parameters, and a refiner with 6.6 billion parameters.
Enhanced CLIP Models: SDXL 1 utilizes multiple CLIP models, namely OpenCLIP ViT-bigG and CLIP ViT-L, to convert and analyze prompts. While CLIP ViT-L is a staple from SD 1.5, ViT-bigG is a powerful new addition that greatly simplifies prompt writing due to its enhanced understanding of human language. I’ll also cover the new way of implementing these CLIP models within ComfyUI.
Optimized Workflow: Lastly, I’ve set up an optimized workflow that minimizes Vram and GPU usage. Now, anyone can easily generate 4-megapixel images with only 12GB of Vram. I’ll walk you through how to work with the base model, refiner, and upscale process.
Join me as I explore these advanced features, providing you with the knowledge to harness the full potential of SDXL 1.