Huge Stable Diffusion XL (SDXL) Text Encoder (on vs off) DreamBooth training comparison

Furkan Gözükara - Oct 9 '23 -

- Dev Community

U-NET is always trained.

All images are 1024x1024 so download full sizes. Each grid image full size are 9216x4286 pixels.

Public tutorial hopefully coming very soon to SECourses (https://www.youtube.com/SECourses). I am still experimenting to find best possible workflow and hyper parameters.

I made a short tutorial for how to use currently shared config files : https://youtu.be/EEV8RPohsbw

PNG info shared in captions of images

For full size images with alt tags please check this reddit thread: https://www.reddit.com/r/StableDiffusion/comments/173qn0b/huge_stable_diffusion_xl_sdxl_text_encoder_on_vs/