The model I think is fairly self explanatory.
So, I guess the new hot thing is finally here, took a year, but then again it took a year after pony for illustrious. I wonder what happened to the illustrious guys, the 3.5 illustrious models were funded fully like a year ago. I guess they took the money and ran without releasing it? Their loss, I suppose.
Anima models will probably go through the same learning phase for me as I learn tricks of the trade on how to do my unique way of doing things again, this one was a bit underbaked I think for replicating the style, but the intention was the concept, and that works fine. Just thought I'd start the new Era of image gen models on an interesting foot by doing something nobody else would probably think to do.
https://vocaroo.com/1dCKgpymuk1M
Back to my devilish tricks.
Works well at a strength of 1
Previews made with Anima Base Model 1.0
Description
FAQ
Comments (8)
Just as i got rid of or replace the pony models, there is a new new thing 😄.
Is anima really that better?
It's apparently better at following prompts and can even use natural language, like you prompt with a sentance instead of just tags. It'll take probably half a year to be a real replacement with checkpoints people make, illustrious has about 1-2 years of a catalog, so it'll take time, the main thing that interests me is that apparently it's better at doing multiple character stuff, with illustrious tags bleed through and it's hard to do scenes with more than one character
@Malebolgia If you think that's good than i better retrain my only lora (and soon a new lora), to it.
Well, Anima is better at understanding prompts, which allows it to generate, with varying degrees of success, things the model doesn’t know via a prompt, without LORA. At the same time, it follows the description quite well when it comes to multiple characters, provided, of course, that the prompt is written correctly, rather than the way DeepSeek usually tends to do it.
Moreover, judging by all accounts, the model works better with tags; it has something akin to an ‘association system’, which allows it to work with tags that have too few images simply by adding related tags.
It also handles backgrounds much better, provided they are described correctly, as well as lighting and angles
But most importantly, it works very well with styles; although it can’t combine them like Illustrious, the comic style actually looks like a comic, rather than an attempt to make an anime character’s eyes smaller and nose bigger, and using the ‘ink brush’ doesn’t clutter the scene with ‘brush strokes’ but rather honestly attempts to give it the look of traditional media.
I don’t know how well it learns when it comes to training Lora, but it may well be much easier to teach the model new concepts or styles than it is with Illustrious.
Anima is a checkpoint?. That i can use on locally?.
@fluffyboi4daddyÂ
https://civitai.red/models/2458426/anima?modelVersionId=2945208
Yes, to run it, you need to use either ComfyUI or Forge Neo.
What’s more, the mod, including everything you need, takes up no more space than the Illustrious mods (even if it runs a tad slower).
@Shikinari ty for info
I'm really happy that one of my two favorite lora creators for comic style has taken an interest in Anima. I hope to see a new lora with a comic artist I might not have known about, but who looks pretty cool, almost every day in the near future.