This is a LoRA for Ratatatat74's style, created upon request.
I will remove this if requested by the original artist.
Info:
Trigger word: @r4t4t4t4t,
Note:
Versions 1.0 and 2.0 are LoRAs trained on nearly identical images using different settings. You might not notice a big difference between them, so feel free to use whichever you prefer.
Description
Trained on Anima Base v1.0.
Trained on artwork from January to June 2026.
Recommended prompt: sketchy outlines,
FAQ
Comments (5)
I think here is the limitation of anima. I like this lora but the lineart is still too perfect. I have tried training this artist a dozen of times, and while it got better the lineart is still not as sketchy as the originals
I completely agree. I actually made this LoRA after someone requested me to recreate Ratatatat's sketchy lines.
I even zoomed in and cropped those sketchy outlines and tagged them as "sketchy outline" to encourage the model to learn the original lines, but perfect recreation was just impossible. Although, generating without quality tags does make the recreation a bit better.
By the way, I saw your Ratatatat LoRA. Did you receive a request from the same person? Or wait, are you actually them? haha
@toghashie441 No, it wasn't me hehe, But this artist is the reason I started to train loras, I like it even outside of porn. I have to say, from the ones I have seem posted here this lora is the closest to the original images.
And it's weird, because you can see the imperfections on hands, clothes and some other parts but they just don't exist on the rest of the body. So it's not that anima can't do that kind of drawings, but something else. I think it could be:
1. That it needs more res on training, my lora is actually trained with 1536x2, but I made a mistake and the trainer didn't see my tags. It got texture pretty well, but that's it.
2. The artist actually has different art styles, even if minimal. The one in your lora is one with thick lines, but they have others with more fine lines. The result is that you get an average, in some images the lines are more clean so it reduces the sketchy effect. This is why I trained mine only on twitter images, which doesn't have lines as thick.
I trained a lokr on Nyantcha (I'll post it later) while fixing the problem I had with the tags, and the result is very good. I made one of ratat but it's not better than the lora I had posted here. Next month I'll train the same lokr of ratat but with 1532x2 and more steps and then I'll decide if I'm giving up or what
@sovietlat I totally agree. I've been experimenting with training since Anima dropped, and honestly, Anima just really struggles with this kind of sketchy style
Huge thanks for your amazing work all this time!