Welcome to magnANIMA!
I love seeing your sample images, it's really encouraging when you post your generated images in the gallery! Thank you <3
This is a new finetune of anima based on my own customized dataset - not a lora merge - and unique from BDD and REanimaTE.
Rich variety in styles, moods, and compositions
Consistently aesthetic and pleasing results with nice lighting, details, and atmosphere
Strong prompt understanding that delivers natural, lively images
Focused on character art, but flexible
If you enjoy anime-inspired art that feels artistic and full of life, magnanima is tuned exactly for that. It keeps the strengths of the base model but adds my personal touch for more creative freedom and visual appeal.
Description
Same dataset as magnANIMA - trained on anima turbo - quite different results!
12 steps
Euler
Simple
Personally I think REanimaTE Turbo turned out better.
FAQ
Comments (11)
This near realism model is great would be better if this model also support multiple views since i like to use it. I hope the dev would consider allow multiple views for this model. Thanks and good job devs.
What do you mean, multiple views?
@CallMeMaybe ah, sorry. its another word for prompt panels. I just figured it after i write this comment. lol
Anyways, thanks devs, after downloading and do trial and errors for the last few days. i found this model is the best model i can found with anima base in civitai. I wonder why this model is so under rated? :(
@scarletaver 'discovery' is hard, there are so many models out there, how is someone supposed to know this one is good? I put these out because I enjoy making them and want to share, but I don't really expect them to get attention. Posting images using the model helps!
Btw dev, do you plan any update for this model? i am excited to see what improvements you may want to came up with this model.
I'm always tinkering, it's just a matter of whether there is a noticeable improvement worth sharing. What would you like to see improved?
oh, i just remember how about increasing type of Chinese girl for update since I only found mostly japan and some korean girls. lol.... don't mind this comment dev if you think my ideas in not worth it. This model is already perfect for me. I just fantasizing about more content. XD
also one more, i just figured this model can't load male kid (shota) or very old man (elders). I know in most model these people almost non-exist but it would nice to add them for someone like immorality like me. @_@
lol.... don't mind this comment dev if you think my ideas in not worth it. This model is already perfect for me. I just fantasizing about more content. XD
I’ve tested both V1.0 versions: the Turbo version and the Base version with Turbo LoRA. I really like both for semi-realism, but I think the Base version with Turbo LoRA is just a tad better. For me, er_sde/sgm_uniform works very well at 8 steps—at 10, it starts to burn out—and if you want a bit more detail, such as more realistic skin, you can change the ModelSamplingAuraFlow setting from 3 to 1.
Thanks for the checkpoint, i like it :)
yeah, personally I'm not as much of a fan of anima turbo as a base to train on. I prefer base + turbo lora.