PDXL Ver
I think it need a little more work, but is pretty good for now! I know what the problem is and for now "simple background" in the negatives is a good enough fix.
This is probably my favorite art style so if you get any good gens please share!
SD 1.5 Ver
Finally got around to fixing this LoRA. Hopefully people like it!
Why doesn't my output look like yours?
Check the negatives in my examples. Do you have those embeds?
I used !ADetailer because it's just a set and forget quality bump.
The images were genned at 512x768 then upscaled 1.5 times using 8x_NMKD-Superscale_150000_G with denoising 0.45. You can find it with a quick Google. Note that this upscaler is non-deterministic from my testing so your images may vary slightly. (Check the bottom of this post for info on the cover image.)
Settings and Training:
Trained on Anything v3
Genned on Anything v4.5
LoRA weight 0.8
Description
Rebuilt entirely.
FAQ
Comments (12)
BASED!
(...on Harada's artsyle)
edit: also i' would like to encourage people to try this at 0.5 to 0.4 on top of whatever they are generating to get some pretty shading without forcing Harada's classic body proportion/faces.
Might be obvious for some but still wanted to write this advice.
BTW, having used your model like a trillion of times i literally cannot stop adding it to whatever i'm generating like "how it wold look if it was from a N1 game,, v2 is clearly superior for the very little i have been using it, congratulations!
@DERESUTE I'm glad you like it! Also, I'm happy to hear it was an improvement! The original was one of my oldest LoRAs, so having gotten better is always good haha
@thesauceer by the way do you know the artist called mota he has done artwork for some "famous"(read niche) videogames and is friend with Harada IRL, it would be awesome if you try to do a lora as good as this one but with that art style, just a suggestion, of course.😃
@DERESUTE I like training art styles and that one seems unique enough, so sure! I'll give it a shot. Hopefully today, but that depends if I have time or not haha
@DERESUTE Oh someone just trained one! If it's not to your liking let me know and I'll take a shot at it after theirs has had some time to be the only one.
@thesauceer btw i was gonna post some stuff but its based on your V1 model, idk if should now, they are very crappy anyway..
@DERESUTE Up to you. I know there is a way to specifically post under one of the versions, but I'm not sure how that works to be honest.
@thesauceer will check, thank you!
works really well. i took the recommended settings, but i have the problem that mine looks less saturated than yours. Any advice ?
Edit: it works perfectly now, thank you :)
If it's saturation it may have to do with the VAE you're using. I forget the name of the one I started using recently after testing a bunch. Give me a moment and I'll get back to you. Glad you like the LoRA!
Alright the VAE I'm using is here. Out of the ones I've tested I liked this one the most, but of course VAE choice is pretty much an opinion thing.


