Hello and welcome to Unholy.
Unholy is a LORA specificly oriented for character design (hard to be flexible for landscapes although didn't test it at full). It brings elements like bioluminiscence and sci-fi but can aim to fantasy due to the trained data.
The main concept was simple: Create Sci-Fi + bioluminiscence element/s/chars but in a certain trained style. This LORA is the result of several trainings, curated selections and more trainings and polishing. It was trained via Noise Offset and Multiresnoise. This is the Noise Offset. If you're curious about the multires noise I will be releasing it on June for our Patreons. And probably free here after some months. Will put an image so you can see the difference.
About the LORA:
- The trigger word is "sc3pt4" and will bring heavier the Sci-Fi aesthetics with the lighting so the values recommended are 1 and 0.7-1 but you can try lower and I suppose it will work.
- The LORA works with and without token. If you find an extreme biasing to the sci-fi you can lower the LORA's strength or you can stop using the trigger word. Both works.
- Side portrait has a different understanding due the training same as portrait. Test it yourselves.
- It was tested on 8 different models but I think Colorful v3 and Rev Animated shines more (again, this maybe cause the seed).
- Of course you can test it but again, this LORA is not intended for landscape generations.
- This LORA works good with other ones, feel free to mix (and play with "Composable Lora" if you find not mixing in the right way).
IMPORTANT!
I wrote an article here and it's like my finetunning diary for the ones interested with releases upcoming, trainings done, exclusive patreon models previews, personal thoughts, etc. So if you want to see what I'm planning or what I'm into, feel free to take a look.
Other than that, enjoy it and if you like this type of content support me and my pal in our Patreon for more quality goodies (and to pay the electricity bill):
https://www.patreon.com/user?u=90661954
Bye!
Description
V1.0 - Noise Offset version
FAQ
Comments (12)
We recommend following the author's sample image and adding SC3PT4 "Style". It makes it more stable.
(I generated about 10 images for comparison, but could not get a good one, so I omitted the comparison image.)
Yeah, you can always control the influence with strenght or without putting the token, I will put some examples of that (and when you go higher) but I didn't save the prompt cause I converted them for Reddit and delete the source ones. 🙃
You know, I couldn't get the results I wanted using this, but this did start adding some freaking AWESOME outfits to my characters, so guess that's what I'll be using it for!
Thanks. The LORA can be used on multiple ways. There is not a staright one even it's an style aesthetic. As you can see it creates "outfits" (energy, plasma, spectral aesthetics into outfits I mean) but in terms of style you can check the examples plus the rest of people who tested it.
Can I ask what was your original intention? Maybe I can help ^^. Other than that happy you're having fun with this ^^.
@LDWorksDavid Well I noticed you got that glowy effect on stuff, I was hoping to get a similar one but with a more fantasy feel, but couldn't get quite that, though it could simply be a difference of models. For ages one main thing I've wanted to get right is neon tattoos that glow, though that on clothing would also be nice. GlowingRunesAI is the closest I found to help with that.
@Yasha277 Umm, I think I know what you mean... I think it's just the model. If you check the samples with Dreamshaper, Colorful, Rev, etc. It doesn't create tattoos but at some checkpoints I saw that behaviour so probably a bias of the Sci-Fi/Fantasy. Sometimes it's hard to know how it will affect one model or other and that's why I train over vanilla SD 1.5 base, for compatibility and to avoid weird generations or misslead ones. Thanks for the review and your comment ^^.
Badassssssss will post more results
I was testing your loras and theyre great but i think this and shag0d are very similar so I keep the other maybe for the next day
Can I merge this in my own model?
EI hi, first thanks a lot for the feedback! if you're interested into something like this for your personal merging I suggest to take a look at Patreon and in some tiers you can request, send data or sample images to train a model/lora, etc. I think its better that way and will be unique ^^.
BTW:
I wrote an article here and it's like my finetunning diary for the ones interested with releases upcoming, trainings done, exclusive patreon models previews, personal thoughts, etc. So if you want to see what I'm planning or what I'm into, feel free to take a look. It's easy to lose this so, here you have.
I don't usually leave a second comment, but it's gotta be said, this LoRA has so far been a godsend since I found it. Idk how to describe what it does besides make things a darker theme (without distorting the art style itself, might I add). Whenever I've made a character that has a darker more evil feel to it, I will often pop this one on and suddenly the results are much better.
Thanks a lot. Means a lot to find these type of comments, for real. For bioluminic, scifi and lighting things is very good IMO.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.