NEW ANIMA VERSION! (don't need the 'nsfwupscaler' tag anymore)

SETTINGS: Trained on anima base, results will vary on other models. Use anima base model suggested settings.

Excited to bring you the Anima version of this lora!

• I used the same dataset/tags as the popular pony version!

• All sample images were made on BASE without any upscaling or other enhancements!

• I REMOVED THE 'nsfwupscaler' tag completely and kepth the typical wd14/danbooru tagging used for the pony version.

• I wrote custom code to fix a cfg issue I found in the training loop of kohya ss and trained locally for 2040 steps with great results!

I feel like this lora is worthy or following up the old one. I'm looking forward to seeing what people come up with.

**The anima version does not have a workflow. I used a custom sd script I modified in kohya ss to quickly generate the sample images. Settings are accurate, used the er_sde sampler and beta57 scheduler (custom added, pulled from comfyui code).

PONY VERSION:

trained on nsfw photos to increase quality for realistic images. the string 'nsfwupscaler' was included in all captions, however this is intended for normal use, not just upscaling.