My first LoRA, and whew - I love this effect! This is loosely based on Alcina Dimitrescu, and other Gothic/Noir concepts, and produces a beautiful dark "patina" on your images; lots of darkness, high contrast lighting, and stunningly elegant women (with big ol'... youknowwhats). It also pairs well with my Textual Inversion embeddings, as you can see from the sample images.
It's a strong LoRA - you'll likely have to reduce the strength based on your prompt, <lora:GothicNoir:0.6> or <lora:GothicNoir:0.5> work well for most cases.
Please consider joining my Patreon! Advanced SD tutorials, settings explanations, adult-art, from a female content creator (me!) patreon.com/theally
Description
FAQ
Comments (30)
Will have to give this a try next to my slime tutorial model 🥰🥰🥰
No keyword?
No, not really needed. Woman, maybe, but it doesn't really matter; effect can be controlled with the strength of the LoRA.
@theally Ok. Thanks! If you don't mind, as far as you work with LORA much, maybe you can explain me what does "Separate UNet/Text Encoder weights" do? In an extension for LORA "sd-webui-additional-networks". It separates Lora weights to "UNet Weight" and "TEnc Weight". I haven't found any iformation or explanation of what those 2 weights do.
@drakmour I can't explain the weights - it has something to do with https://towardsdatascience.com/unet-line-by-line-explanation-9b191c76baf5 - but I don't use that plugin, sorry. Maybe look at the Python and see what's going on behind the scenes?
@theally yeah I saw couple articles about UNet but it's too complicated and overwhelming for my humanitarian mind. :-D I hoped that somebody who understands what do they do explain me in simple words. Like "If you put 1 here, it will do that". :-D
@drakmour I've done some quick testing. So a LoRA can have both TEnc and U-Net modules, but when you're training you can specify to train with the textual encoder (TEnc) only. This is apparently what produces the textual inversion effect - according to the docs. I guess it's to do with how closely the LoRA image captioning is taken into account vs the images - maybe? Who knows! Someone needs to explain it like I'm 5 :)
@theally ahahah, I've seen a prompt for ChatGPT about "talk to me like I'm 5y.o.". Funny coincidence. :-D I spent all day sliding those 2 sliders from -1 to +2 and comparing the image. I guess I will have a bit clarity when I'll make a matrix like X/Y plot to see the patterns of effects that they make. :-)
Looks really cool! but I'm a bit confused, about this Lady D lora in the descprition while this lora is called Gothic Noir. Is there a place where I should get this Lady D one?
Thanks!
@Redbird that's my fault - I should have called the LoRA GothicNoir, but it's based on Lady Dimitrescu's likeness/aesthetic, so that's what I called it while training - same thing :)
@theally Aah ok that makes sense now, well that's great since I love Lady D haha
Keep up the good work, I really enjoyed your previous work too, the style petal one is one of my favs
I've actually just renamed the model and tweaked a few metadata options. Same model, different filename to call it in a prompt - keeping things simple!
@theallyVery nice good idea thanks!
Also not sure if it was just me, probably was, but my before and after images weren't changing with this lora, but then I updated my webui and It worked very well!
I'll post some images later!
HA, didn't know you could put gifs in there, that scared the shit out of me, good job. :D
Ha! I think I'm the first to do it - I've been talking with Civitai about allowing webm and mp4, but their image processing isn't set up for that (yet), so - crappy framerate gifs it is! I'm using Depthmaps to generate the motion.
Yup, just subtle enough to be jarring. I hope people don't start making neon flashing booblets for their first images, tho.
AHahah, same! I was scared too. I though that I oversitted too late and my mind started to swam. :-D What do you use for the animation? Is it just pulling one area around or there is a parallax effect?
@mousewrites It's gonna happen - I'm already working on it!
@drakmour There's a swing preset in the depth extension.
Good Christ! I nearly shat myself when it started moving! lmao
I'm hearing that a lot, haha! :)
me too, what the hell D:
Nice one!
Sooooo... how does one accomplish that?! In Stable Diffusion?!!!
Well, I'm definitely doing something wrong.
I've tried using the prompts with various values such as <lora:gothicNoir_gothicNoir:1.0>. My lora files are located in \stable-diffusion-webui\extensions\sd-webui-additional-networks\models\lora
I've also tried activating them using the additional networks extension in A1111.
But no matter what I do, there's zero effect. A strength from -1 to +1.5 has zero effect whatsoever.
Can anyone suggest what I'm missing?
Very odd! So if you're using the additional networks, then yes, it needs to be in \stable-diffusion-webui\extensions\sd-webui-additional-networks\models\lora otherwise they go in the \stable-diffusion-webui\models\Lora folder - I'm sure you got that part!
Sorry to sound condescending, but have you clicked the "enable" button on the Additional Networks pane? Also, if you run a prompt, below the image where it shows the generation parameters, it should show AddNet Enabled: True, AddNet Module 1: LoRA, AddNet Model 1: gothicNoir_gothicNoir(7214f6a81197), AddNet Weight A 1: 1, AddNet Weight B 1: 1 if the LoRA is being implemented in your prompt - do you see this?
Lastly, do other LoRA files work?
@theally Actually, no, no other LORA worked. Or at least not noticeably, so I was always pretty unimpressed.
No idea what the deal was with the Additional Networks (and yes, did enable), so I instead just shipped all of them into /models/lora and it works great.
So...thanks for making a LORA that was cool enough to bother chasing this down.
@meritrashgmailcom Oh, perfect! I'm glad we got it working, because there are some fantastic ones out there! If you make something cool with Gothic, I'd love to see! Thanks x
any place I can find ladydimm:0.5 or midjourney files?
Ah, LadyDim IS the Gothic Noir, I just renamed it after I made the images. As for the MidJourney, it's a Hypernet - I can't recall which one, but it's uploaded on Civitai somewhere! Have a search, it will be the most popular one.
SDXL?
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.