If you enjoyed this model, why don't you check out my other works? I put a lot of effort into correctly recreating artist styles and making the occasional character LoRA.😊
You can also buy me a Coffee ☕ at https://ko-fi.com/stickyricky
This model was trained using the collab/guide by holostrawberry [Guide] Make your own Loras, easy and free - colabs | Stable Diffusion Other | Civitai
There already exist other LoRAs for this style on civitai and huggingface that tend to be heavily overbaked or were trained for some other checkpoints. Check the recommendation below.
About the artist:
Pixiv Fanbox: ビモ｜pixivFANBOX
Twitter: ビモ🐙🦑 (@JT_Veemo) / Twitter
Patreon: JTveemo | Creating cute stuff and NSFW | Patreon
JTVeemo is in large part a hentai artist. They are mostly known for making horny fanart of Splatoon and other videogame characters (Pokemon, Zelda and Genshin just to name a few).
Most characteristic are the eyes and a general softness/thickness in design.
Weights & Prompts
Note that these models work particularly well for NSFW images. For inspiration check the examples.
The only necessary prompt is:
jtveemoLoCon (NAI):
The checkpoint used can be found here.
a1079602570/animefull-final-pruned at main (huggingface.co)
This is the goto version for most checkpoints. Unlike the first two versions it was trained and tested on NAI and has far more vibrant colors.
Use a weight of 0.7-1.0, adjust to taste and depending on your checkpoint.
Note, that the resemblance also scales with the number of sampling steps used.
Beware the occassional inklings, which tend to be tied to specific prompts like gradient hair.
More sampling steps or additional negative prompts can prevent that.
AOM3 Lycoris:
This version was specifically trained on Abyss Orange Mix 3. It is (hopefully) the final and definitive version, as it exceeds the others in flexibility and accuracy.
Use a weight of 1.0, adjust to taste. This model is really flexible and is compatible with other LoRA (character or pose models) at full weight. As usual I have included an XYZ plot for reference.
LoRA:
I recommend a weight around 0.7 which is compatible with most character lora. You can go higher if this is the only lora used.
Lycoris:
Use a weight of around 1.0, whether you use other LoRA or not. Adjust to taste.
The Lycoris model is a bit weaker than the LoRA but that makes it more flexible.
Please leave a review if you enjoyed this model or have problems with it. This helps me with improving future models and finetuning them in general. Give it a try and throw some images down below! 👇
Description
Lycoris dataset trained on the same dataset and NAI. Slower training makes it more flexible but a bit weaker than the LoRA trained specifically for AOM3.
FAQ
Comments (16)
Thanks, this looks great. Do you take suggestions for any artist or are you looking for ones from specific genres?
Let's just say I'm not a fan of furry art.
Apart from that I can't really say that I'd prefer specific genres. Just throw them in. If they look cool or interesting to do, I'll give it a try.
@StickyRicky No worries, not a fan of furry art either. Hmm the only other artist besides Kyockcho I have in my top is Nunnu (Zero, Miraichizu) https://danbooru.donmai.us/posts?tags=zero_%28miraichizu%29
🤣 Did you get spammed or something? There like 4 comments all within a day with the same template and two accounts are like 1 day old.
@CanIgetAWutwut Not really. Honestly expected a lot more responses so I could pick and choose :D
Nunnu should certainly be possible. It goes on the list ^^
In case you haven't noticed. the model for Nunnu is done. I'd appreciate some honest feedback/review.
It's really hard to improve when even my 1k+ downloads models get 0 reviews 😒
@StickyRicky Oh hell yes, thanks. I didn't catch that it was posted. I'll make some gens right now.
"I'm slowly but surely running out of ideas for Artstyle LoRAs so I'd be happy to get some suggestions."
I can throw in a few suggestions for you (mostly SFW if you don't mind doing those):
- "Loathing" (from the Kingdom of / West of / Shadows over Loathing - games)
- Don Lawrence (for example from his "Storm" comics)
- Dick Bruna
- Richard Scarry
Don't have a problem with SFW artstyles. The greats just tend to be included in stable diffusion or there already exist models on civit.
Everything but Don Lawrence I probably can't train with my usual methods (model training on NAI (anime diffusion) and testing on AOM3) so those would have to wait until I find a fitting checkpoint.
Thanks for the ideas. I'll see what I can do.
"I'm slowly but surely running out of ideas for Artstyle LoRAs so I'd be happy to get some suggestions."
What about Sparrow
That's gonna be a no from me. To create models you need cohesive art and large sample sets.
"I'm slowly but surely running out of ideas for Artstyle LoRAs so I'd be happy to get some suggestions."
Would it be possible to do something in the style of Syoyo or Yoshimiyoshi?
That looks fun. Not sure how well SD will handle the characters with the backgrounds but at least yoshimi should work. That goes on my list ^^
Small update, the model is done and on Civit. I'd appreciate some honest feedback and if you would drop a review 💖
"I'm slowly but surely running out of ideas for Artstyle LoRAs so I'd be happy to get some suggestions."
I'd love to see an Iruma Kamiri/Hellabunna style if you're interested in making one.
That might be something you can replicate with the right prompts.
Tags | Danbooru (donmai.us) - some of them can massively change the artstyle towards this more oldschool style. If that doesn't work I can take a look.