Share and Review: Please share your generations to this model page and write a review as it will help me understand how to create better LoRA's in future versions.
Illustrios version of my previous lora, trained on the Nova Furry model. Tested on NovaFurryIllustriosXL v3.0 and NoobRealXL. In general, it works normally. But I'm not so good at Illustrios models, so I can't give "better" results, maybe someone else will do better.
I'm uploading 2 versions, because I can't figure out which one is better, although the results are about the same.
I'll call BF and BT (Better Form (e) and Better Texture).
LoRA works smoothly from 0.5 to 1 strength.
If you want to get the result as in the preview, it is advisable to prompt like this (red knot, white shaft, red tip)
Description
Illustrios version of my previous lora, trained on the Nova Furry model. Tested on NovaFurryIllustriosXL v3.0 and NoobRealXL. In general, it works normally. But I'm not so good at Illustrios models, so I can't give "better" results, maybe someone will do better.
I'm uploading 2 versions, because I can't figure out which one is better, although the results are about the same.
I'll call BF and BT (Better Form (e) and Better Texture).
LoRA works smoothly from 0.5 to 1 strength.
If you want to get the result as in the preview, it is advisable to describe (red knot, white shaft, red tip)
FAQ
Comments (10)
I would love to see the actual generation data for the images; because this looks fantastic so far.
I forgot to add, I'll add it now) I have to rewrite it manually from comfi
Added promps, so ypu can check. Also you can take whole ComfyUI workflow from image, and it's also published on furry diffusion discord server.
Your LoRAs by the way also helped to add generations to the dataset x3 I am very grateful to you
@SugarOuroboros I had to hand draw/paint a lot of sheaths with specific form to stop the AI from making them just spheres lol; so glad people are getting use out of it.
@pihlawrkr738 Well, there is partly the same problem with the scheath a bit, I will need to make a second version as soon as I collect enough high-quality data.
@SugarOuroboros Yeah Illustrious REALLY wants to make them balls for some reason.
@SugarOuroboros Do ya got access to 'data' in the forms of img/vid of the real thing? Or you only seek furry? Honestly, this is as close as I've seen to real so far in terms of AI.
@pihlawrkr738 May depend on the model. I think the thrillustrious model has insane potential for super realistic furry, but it's got like, zero training on furry content. I just tried to run a furry lora on it and it gave me weird small, large-eyed feral verions that looked very realistic, or just regular-eyed animals that bared a likeness to the character. In any sense, it looks amazing, it just isn't anything close to furry/anthro.
@Lazman The training data is based on furry art, most of it is generated, I am still collecting additional art for a better version of this lore. As a rule, I took those arts where the penis had the most aesthetic shape, the root of the knot was visible, plus the veins on the penis were in the form of a vascular network, and not subcutaneous protruding veins, like on a humanoid one.
I don't quite understand your question. Is there a problem that the lore sometimes turns people into furries?
@SugarOuroboros Question? If you're talking about my response to the other guy, there's one illustrious model called 'thrillustrious' (no longer available on civitai), and it's a realism model, but it can't make furries cuz of the lack of that in it's training data. But that's THAT model, nothing to do with what you've made.
I understand now though from what you've said, that you just have a very good understanding of canine anatomy. Moreso than me for sure.
But, you may want to put a rush order on that(next version of the Lora), cuz from the recent anouncement on Civitai, it looks like they're gettin spanked by legislators, which is causing them to clean house on anything viewed as 'politically incorrect'.