This "LoRA" was trained using the output of "NovelAI". (based on WD1.5)
It can be used with "WD1.5" to create an animated style.
Description
FAQ
Comments (7)
I'm running into an error with this one, is it still functional? Trying the samples, for examples, I get:
output with shape [128, 320] doesn't match the broadcast shape [128, 320, 128, 320]
SD2.1 768?
Not SD 1.5.
If it is SD2.1 768, please apply from additional net or lyco
where can I find WD1.5,give me a hint,plz : )
Doesn't WD already give an animated style?
The quality was very low at the time of this release.
If this model had initially been trained on Anything v3, it would have been a little more successful.
There is no strong reason to make such change now, because NovelAI released its first models to the public.
https://blog.novelai.net/novelai-diffusion-v1-weights-release-en-e40d11e16bd5
But I invite you to train a lora for Anything v3 (SD 1.5), Animagine (SDXL), NoobAI, Illustrious, or Pony, intended to emulate the style of the second model of NovelAI.