How exactly did you train your model?
#2
by
Rulk
- opened
I used TheLastBen's Fast Dreambooth Colab. 5200 Steps. 17% Text Encoder. 32 Images. No regularizations. Images consistent with each other and in good quality.
Hi, thanks for advice!
I was able to achieve good result with regularization images generated by stable diffusion with prompts corresponding to depictions on training images.
@Rulk might want to check out scenario.gg for game assets
artificialguybr
changed discussion status to
closed