Hugging Face's logo

Gazingstars123
/
Anima-2.9B

Text-to-Image
Diffusion Single File
English
anima
comfyui
Model card Files
xet
 Community
18

A column of slightly stupid questions from an inexperienced lora expert

#16
by Shadyman961 - opened
Discussion
Shadyman961

Hi!
First of all, thank you so much for your work. You're awesome! I only use your Anima trainer to teach lore!
I hope everything works out for you, good luck!

Now for the questions:

  1. I'd like to clarify regarding the Lore style. Will the styles from Lore be applied better to characters trained in Lore? In the basic Anima version, when I implemented any checkpoint other than the basic one, the original style would start to fade and barely change to the Lore style... I have a lot of characters that I'm training, and I'd like to apply different drawing styles to them, etc., but Anima doesn't always allow it....
    1.1. Will the extended Anima be able to mix the styles of artists already included in the base? For example, kairunoburogu + nyantcha, or any others of your choice?
  2. Do you plan to support the extended version in Inpaint or Forge Neo?
  3. I've seen a lot of questions about custom text encoders and Vae. Will you use a text encoder and Vae different from Anima or keep the default ones?
  4. Will you try to fix the upscaling issue with the default version? (As far as I remember, Anima in the base version can be upscaled a maximum of 1.5x with hi-res fix, unlike Illustrious, which can upscale by 2x and 3x.)
  5. (Sorry, but this is very interesting, off-topic.) How can I train a model so that the Lora character style matches the Lora style?

Thank you very much for your time, and good luck, you'll succeed!
Gazingstars123
 Owner
  1. If i understand your question correctly, the fix is on the training side, not the model side. Character LoRAs need style-diverse datasets (as many different artists as they can find for that character) and explicit artist tags in the captions. Also: stack at ~0.6–0.8 weight rather than 1.0.
  2. It's already supported natively in Forge Neo
  3. I won't say no, but it's a not an easy job, and the capability to actually train them depend on two things: time and money. So it can either takes very long or very fast depending on people, and if they are willing to support the project. And it's something that has to be decided today, not later.
  4. Upscaling works a bit better, but images do start getting noisy at around 2.5MP or above
  5. Sorry man, I don't understand this question

Sign up or log in to comment