XL-v2 Based on Pony Diffusion V6 XL
LoRa trained on some gothic style arts.
All versions here: https://huggingface.co/We11/Goth_girl-v1/tree/main
Also, if you want to create high-resolution images, I recommend using "ADetailer" for the good looking faces.
Setting i use:
for XL-v2: 832x1216 No Hires. Sampling steps: ~25 Sampling method: Euler a CFG Scale: 7
Can be used with no other LoRa, but sometimes i use with:
gothic neon v1.0 ~0.3
[GP] somethingweird style [Pony XL] ~0.2
Concept Art Twilight ~0.8
And Vae SDXL-VAE
for SD1.5: 512x768 or 768x768 or 1024x1024 resolution. Sampling steps: 27 Sampling method: Euler a Upscaler: Latent Upscale by: 1.5 CFG Scale: 5.5
SD1.5 v2 trained at 1024x1024 resolution. There are 2 versions that I use to generate, regular v2 and v2S. The regular v2 version is more stable, and v2S is more “Stylish” with more details and the style of the arts on which it was trained on, but less stable especially with faces. So you can try both if you want. Or you can still use v1.5
Description
FAQ
Comments (26)
For those who generate using CIVIT: Data of some picutes is scuft for some reason, it doesn't show Concept Art Twilight and puts [GP] somethingweird style [Pony XL] 's value at 1.0 when its 0.2 for me.
Be careful with it if you use "Remix"
hi, could you help me? How do you create such beautiful Loras, which look a lot like the anime? I tried some: :hollowstrawberry-kohya-colab
But I get Loras who don't look much like the character I wanted to create :(
I feel like the key is in building a solid base of images featuring your character/style you want to replicate. The more high-quality artwork you have, the better the results will be. It's also important to capture the right prompts that you want reflected in the final results. For this, I used "WD14 Captioning" in Kohya_ss. If you're training in Kohya, try creating directly on CIVIT and compare the results. It's quite simple. Most settings are already well-configured; you only need a base of images. Or watch a video about what settings to use in Koya, there are a lot of them on YouTube
Besides, even if you use the same settings and the same database, you can still get different results. So sometimes you just gotta try again
@We11 I only find tutorials of old versions, while the new version is totally different, so I don't know what to do :(
@ApexThunder Try training on CIVIT and then copy the settings :D
hi,i have a weird issue where he only do close up, any idea why?
Depends on the prompts you use. I usually use "Portrait", "Upper body" or "full body" if I want to see certain results. But if this still happens even with these prompts, try highlighting them with brackets like "(full body)". Also give me an example of your prompts pls
@We11 ya you are right, i used to many prompt for the face lol
really good lora!!! i love how it combined with my lora and made the image come to life!!
I'm definitely guilty of overusing this, but it works in anime and photorealistic forms really well, and is my go-to Goth waifu
I'm glad you liked it!
Used it make my first images, top tier model.
If applying the gothgirl lora removes your character's eyeshadow then there must be an issue somewhere...
Not all Loras are compatible I guess
Beautiful
So, I was working on a 3.0 and uploaded some results to get feedback, but the images just never get anywhere and no one sees them. Meanwhile, the guy who makes 10 identical pictures with "tentacles porn" gets into the feed every time. And I often see interesting and unique ideas that people come up with in their work, but no one sees them, which makes uploading them pointless. I don't know how the algorithms work, but I know there's no point in uploading to this site. 3.0 won't be here.
Sad seeing this very good LoRA, that i have used a lot when i used Pony, not getting enough recognition. T_T
What site are you going to switch to?
@Sadara Im not actually switching to any website yet. I only have a test version of v3 rn, but if i finish it, I’ll probably just upload it to Hugging Face for those who interested
Uploaded v3b test version, which I'm currently using for Illustrious at Hugging Face
@We11 Thank you some much! 💕
Does it work with illu?
Hugging Face v3b is for illustrious
Pony one is also works tho
great!!
thank you
Details
Files
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.