Update
Added 2024-03-06 20:00:06 +0000 UTCI have been testing a SDXL mix that I have made. You can see some images on my Twitter.
What's the difference you may ask?
The SDXL model is very good with basic prompts and does know a lot of characters without needing an extra model (Lora) of that character. It is also more capable in terms of generating anatomy, although still not 100% perfect.
I had difficulties with the speed of SDXL model. It was around 5x slower per image, but I found a way around it by using a different (more efficient) UI. I still need to tune the style of the SDXL model to my liking.
For the future you may see a mix of SD1.5 and SDXL generated images.
