TheLastBen / fast-stable-diffusion

fast-stable-diffusion + DreamBooth
MIT License
7.41k stars 1.28k forks source link

fast-DreamBooth Colab not creating good results compared to Runpod #704

Open afrofail opened 1 year ago

afrofail commented 1 year ago

The google collab is more convent and has been easier to use, however the runpod tutorial I followed by Aitrepreneur has drastically different results that are working way better for what I'm trying to generate. I trained a character in both and the fast-DreamBooth collab results just the character with no other prompts. The runpod blends the characters face with prompt and style flawlessly. Is it because it's a 1.4 model?

I'm doing 20 images, 500 step increments, at 1000 steps it's already not implementing other prompt keywords unless I turn down the cfg scale to 1-3, but as soon as I hit 1500 steps, it's hard to get any keywords in unless I'm doing image2image. Am I doing something wrong or should I convert back to 1.4 model?

0.00001, 50 train text encoder.

TheLastBen commented 1 year ago

set the text encoder to 10% and do 6000 steps

afrofail commented 1 year ago

I'll give it a shot and see if this helps. Thanks for the input and keeping tabs with issues all the time, I know it's really hard to keep up with everyone and there personal issues.

0xdevalias commented 1 year ago

See also: