Simile = comparison, likeness, parallel, similar, resembling
Note: Preview images are generated without the use of LORAs on ForgeNeo
Anima ver.
About
[Version v1.0 - 17/6/2026]
This is my first attempt at an Anima-based checkpoint model, and it is very different from SDXL models, so I'm still not sure how it'll perform for other people. This is based on the initial full release of Anima base-v1.0 model, so the same licensing and permissions applies. As usual with all of my checkpoint models, I try to make them more male focused, specifically with bara body types. But you can still get twink and bishounen characters with the right prompts.
And for people who are new to Anima models, the prompting is pretty similar to Illustrious XL and NoobAI models. The main difference is Anima is better at handling natural language prompting than ILXL/NAI (see examples). It also uses primarily Danbooru and Gelbooru-style tags. Unlike NoobAI models, they don't use e621 datasets, so their tags don't work. Artist style tags works similarly to ILXL/NAI models, but you have to add @ in front of the artist name. So instead of dizdoodz, it has to be @dizdoodz. Character tags work the same way with ILXL/NAI.
As I have mentioned earlier, this is my first attempt, so I'm still very inexperienced with Anima. If you encounter any issue I welcome any feedback and I will attempt to improve and refine them in future updates.
Tips and Usage
Unless you're using Civitai's generator, you'll need either ComfyUI or ForgeNeo to run it. Check the official Anima model page [HF] for the required text encoder and VAE file that's necessary to run it.
My Settings
Resolution: between 512^2 and 1536^2
but my go-to resolution is 1024x1536
Sampling Method: - ER SDE (my go-to default sampler)
- Euler a (cleaner and smoother, better for anime styles)
Sampling Steps: 25 - 35
Upscaler: R-ESRGAN 4x+ Anime6B
CFG Scale: 4 - 6
Hires fix: upscale - 1.5x
upscaler - R-ESRGAN 4x+ Anime6B
denoising - 3.5 - 4Note: Do NOT go over 1.5x for the upscale or your image will get messed up.
Positive Prompt:
masterpiece, best quality, very aesthetic, extremely detailed, score_7Negative Prompt:
worst quality, low quality, lowres, score_1, score_2, score_3, jpeg artifacts, bad anatomy, watermark, artist name⚠️DISCLAIMER⚠️
Please do NOT generate illegal content with this model. You are solely responsible for any unethical use of this model(s)
Model License
Description
FAQ
Comments (48)
Finally ♥️♥️♥️
i'll send some men on your bed 🤣
They better be top-tier dilfs lol.. 🤣
This model is pretty solid. It handles the basics you'd expect it to, and I like it.
But I do need someone to teach me if it's possible to mix styles, or if I should just stick to using one artist tag at a time. I did try mixing them, but it got quite messy. And sadly, some gay/yaoi/bara artist tags don't work very well - maybe because there aren't many samples of them on Gelbooru?
Anyway, I do have some LoRAs to help achieve the look I want, but you did such a good job with the old NoobAI model in this regard that it kinda spoiled me rotten. I’d prefer not to rely on LoRAs too much, or not use them at all, lol.
I need to test it more, though. Good job!
Just checking, but did you add '@' in front of the artist tag? If not, the effect will be very weak. I will say though, even with the artist tags, if you want to get the best result for specific artists, I think its better to just use a LORA. For me artist tags are just to add some 'flavor' to your gens lol..
Thanks so much for this! Could be a sizable step forward for Anima based on just how many popular Illustrious merges use Homosimile (and for good reason). Can also attest that furries work just fine on V1!
Time to celebrate — HomoSimile now has an Anima version too~~~~ It's time to join the Anima
That's amazing, I've been waiting for it for a long time.
Finally a good Anima model, this is delicious. Thank you Kairen!!
We have been blessed again but this time with an Anima model? Lol, I'm gonna take my time with this one. Thanks for the always amazing work, dude!😍❤️😘
This model is marked as a "Checkpoint Merge"; could you provide details on this? Is it a LoRA merge rather than a finetune?
Thanks for your efforts as always!
Both actually, it was finetuned and then I added some lora as well.
@Kairen92 I see, thank you!
is there a list somewhere of trained artists?
For anima models I haven't found one yet, but there's this list for ILXL/NAI artists. It hasn't been updated for a while but some of them probably still work for Anima as well. Just make sure you tag them properly with '@'..
I've been using https://animadex.net/?mode=artists as a reference for artists
Really man? K.
Huge fan of your work, and excited to see an Anima release, thank you! ❤️
Like your previous models, are there any artist tags that were specifically trained into this model (e.g. ppatta)?
Not yet unfortunately. Since it was my first attempt at an Anima model, I'm focusing more on making it more compatible with bara and muscular body types. Artist tags also works a little different from Illustrious/NoobAI so it's gonna take some time for me to learn and experiment with. But I am planning on doing that for the next update..
@Kairen92 I've loved every model you've released so far. I'm excited to see how this one evolves❤️
@meer anima base can do gay stuff perfectly. for artist names, they always start with @ like @artist name. You can use pony-style tags score_9, score_8, score_#. ... if you want quality tags
The checkpoint I've been waiting for :)
I've noticed that using the 'R-ESRGAN 4x+ Anime6B' upscaling model results in a significant loss of detail, to the point where the upscaled output actually looks worse than the original image. Could you suggest a way to mitigate this, or is there a better alternative for anime-style upscaling?
Hmmm, have you tried 4x_NMKD-Siax_200k or Remacri? Maybe you'll get better results with those. Just make sure you use the correct settings for them..
Could we get a workflow for this model? I'm new to ComfyUI and don't really know where to start. Awesome work as always~
Hi there, I don't use Comfy that much, so I can't really help you with that. But from what I heard, you can just download any images you found on this site that uses Comfy to gen them, and just drag and drop them into Comfy to get the workflow.. Just make sure they're for Anima though..
THE BARA GOAT HAS JOINED THE ANIMA MOVEMENT!
Could the limb stability be improved in the next version, especially for the hands and feet? Anima's hands and feet always give me a really hard time. 😭
Yeah, initial Anima release is pretty wonky when it comes to anatomy and limbs. That has been one of my main things to improve alongside making it male-focused. Hopefully it'll be even better in future versions.. 😅
What VAE are you using?
Also, are you using ComfyUI or ForgeNeo for the sample images?
No VAE, just the mandatory one for anima models, qwen_image_vae.safetensors..
I used ForgeNeo for my gens..
@Kairen92 When I generated a NoobAI model with reForge, it took about 1 to 2 minutes to generate a 1.5x size (1500x1500), but when I generated an Anima model with ForgeNeo, it took about 5 times longer. Is this normal for everyone? Or is there some trick to generating models with ForgeNeo?
@yuni623 Well, anima is slightly slower compared to IL/NAI but not by much. 5 times is quite a lot. For comparison, when I gen 7 images with NAI its around 3 mins, but with Anima, 7 images with the same resolution is around 4 mins and 30 secs. I also use ForgeNeo with minimal extensions and settings adjustments, everything is basically the default settings. So if your Anima gens are way too slow, it might have something to do with your graphic card, maybe?
Man, this anime checkpoint is a game-changer. Great prompt adherence and it knows famous characters and styles perfectly. What really blows me away, though, is how it handles different elements—especially text inside the image! It’s usually a nightmare with regular checkpoints, but here it works so well.
Thank you for your kind words, I'm glad you like it.. 😘
Is there no way to train this model using LoRa? An error is prompted during the training process
I wouldn't recommend it. It's better to just train on base Anima..
May I ask about the number of images used for training?
Not a lot, about 3000+
Ex: In the past, I really got out of generating images or anything to do with this website, but I feel Anima has given me a reason to start trying to generate again. I really love this model and Anima especially since it allows more detailed prompts, stylized forms and more dynamic depictions than illustrious models. The only problem I really have is with the hands and their movements but other than that I think this is one of your best models yet. I’m really excited for any future updates to Anima or new models you may make going forward because what I see now is already incredible. Great work dude. ❤️👍
Thank you so much, I really appreciate that.. ❤️
And yeah, I get what you mean. I was already planning to retire from AI (at least from Civit), but then Anima comes along and pulls me back in.. Now I'm itching to play around with it more.. I already have so many things planned for the next update, I just wish I have more time to work on it.. 😔
@Kairen92 Nah I feel you. Half of my post on here were just scheduled backdrops because I also have no time or interest in it right now. However, I do wish you the best.
I really, really like this. I hope you can keep going.
That's the plan, lol.. ☺️