V2 ON BOOSTY and REALISM SLIDER FOR ANIMA
ANI-RL is an experimental semi-realistic model based on Anima (base-v1.0). This is the first version of a personal project aimed at combining Anima’s strong artistic understanding with a more grounded, semi-realistic aesthetic.
Simple Use (Load Checkpoint): All components (DiT, text encoder, and VAE) are included in a single file for convenience.
Key Features
Semi-realistic style with good anatomical structure and lighting
Maintains Anima’s strong prompt adherence and artistic knowledge
@ANIRL trigger significantly increases anime-like stylization when needed
Works well for both semi-realistic and stylized outputs depending on prompting
Recommended Settings
Sampler: er_sde (or other)
Steps: 20–40
CFG Scale: 4–6
Trigger Word
Use @ANIRL at the beginning of the prompt to push the generation toward a more anime-influenced look.
Without the trigger the model leans more semi-realistic. With @ANIRL it becomes noticeably more stylized/anime-leaning.
Example Prompt Structure
Positive: @ANIRL. masterpiece, best quality, score_8, highres, absurdres [your description],
Negative: worst quality, low quality, score_1, score_2, score_3, score_4, blurry, deformed, extra limbs, bad anatomy, artist name
Version Notes
v1 — First experimental release
Focused on achieving a balanced semi-realistic look while preserving Anima’s core strengths
Future versions will improve consistency, lighting, and fine details
This is an experimental version. Feedback is welcome.
If you enjoy using ANI-RL, please:
👍 Like the model
📤 Share your generations
💬 Leave a comment
💡 Suggest ideas for future versions
☕ Support development via Boosty
License & Usage
Notice: This model is protected by copyright. Copying, distribution, merging with other models, or use on other websites without the explicit written permission of the author is prohibited.
Description
FAQ
Comments (22)
Thank you for making a Checkpoint with all the components included. I love your models.
thank you!
This is a truly magnificent checkpoint. Yeah yeah porn xxx boobies good. The creativity is awesome. The checkpoint has a built-in bias towards shallow depth-of-field that works really well for portrait scenes.
The style is incredibly beautiful, one of the best I've seen! but for some reason I have problems with my hands (is it just me? Am I doing something wrong? The fingers turn into a wild horror(
In my experience, this is a more general problem with Anima. Hand quality is poor overall and a lot of the hand detailer models that I use in workflows make things even worse.
Just an anima problem in general - some loras have better stability for hands, I normally combine at a low weight with them if it's an issue.
I tried literally terabytes of different models. But this immediately became my top one. Extremely well follows prompt. And the way it handles lighting is amazing. And workflow worked flawlessly. I tip my hat before this mastery. Big thanks.
Thank you so much for your feedback and pleasant words.
This is amazing. Easily the best Anima model right now. Keep up the good works!
this model looks amazing. do you think you could do an anima version of your LDSCH lora its the best lora ever.
yes, of course. soon I will be training styles for the Anima model, now I am studying and experimenting.
Amazing fine-tune, it was really fun to play around with it! You’re seriously talented 🔥
It would be awesome if you made a fully 3D Disney-like fine-tune without the anime aesthetic. That’s a huge problem in the open-source community right now 😅 There’s basically only one really good Illustrious fine-tune, and that’s it. The other 99.9% are anime fine-tunes 😁
A day, i've throw all FLUXSDXLZ away.
P.S where i can get guide for BSS Anima Loader Sampler and Casher? Want to improve generation speed.
@llikswonskcalb спасибо.
Same
gorgeous model, I have only one problem, the shadows are too strong, I can barely see the details behind the shadows. using loras or adding negative prompts doesn't really help. I don't know, maybe I'm missing something, any advice?
P.S. It seems like this only happens to male characters with certain attributes, the shadows are so thick, I can't even see their faces, no problem on female characters so far.
I sometimes have the opposite problem, where the characters in the foreground are almost hit by flash photography or something. Wish we could balance it out, or maybe there are some keywords we're both missing. Turns out it was the turbo lora.
Overall I really like this model because of the flexibility with the lighting. Many other models will always illuminate the subject regardless of prompting - this one is very flexible.
@ANIRL. keeps popping up in images at times.
Sometimes the model thinks that ANIRL is not a trigger word, but the text that should be on the image.
After trying out about 50+ anima checkpoints, I consider this one of - if not the- best checkpoints still for anima: 1) It does lighting well - many other checkpoints will always have diffuse lighting in the foreground regardless of prompting. 2) It easily combines with other artist tags and loras while still keeping a high quality feel - extremely realistic or cartoon are easily achievable. 3) It can render at multiple different viewing distances and angles. 4) It handles a variety of prompting well - it already has many concepts baked in that other checkpoints struggle with. I think for a next version the following could be addressed: 1) Skin condition prompting - without explicit prompting for each region it struggles with changing the skin condition (dirty, muddy, scratched, blemished, etc). This can be circumvented by prompting for each region - dirty cheeks, dirty breasts, dirty arms, etc. 2) More dynamic action - similar prompts in different checkpoints will offer a more dynamic or motion driven image. Think surprise tentacle attack, accidentally dropping an object, or getting dragged through a forest. Overall I consider this to probably be the best checkpoint currently for anima for my style preferences.