Update Flux Version
The new version has been trained to work on Flux models. From my testing Flux is amazing with sentence like prompts, and turns booru prompts to anime pics that don't have a lot of details but this LoRA fixes that:
I tested it on SD Forge using the NF4 model and it work fine, you just need to change the Diffusion in Low Bits from Automatic to Automatic (fp16 LoRA) like this:
The rest of the LoRAs are designed to work with Pony Diffusion XL, or any other Pony model.
How to use
1- Manga Style (Blood Moon): Made from high quality monochrome images to give you a manga feel, oh and I called it blood moon because if you have something red in your prompt you might get a cool effect xD.
Trigger words: This LoRA is without Trigger Words, but having "monochrome, greyscale" in your prompt can greatly enhance the results.
Works on: 0.6-1 weights
2- Anime Illustrations Style: Made from very colorful and high quality data to improve your images and give them a badass feel.
Trigger words: MythAn1m3
Works on: 0.6-1 weights.
3- Portrait Style: This style is made to create high quality portraits of fantasy characters, and most of the training data is made with semi realistic art style.
Trigger words: MythP0rt
Works on: 0.6-1 weights.
4- Flux: This style is made to create high quality fantasy art, and it has training data similar to the Portrait Style
Trigger words: MythP0rt
Works on: 0.5-1.5 weights, feel free to experiment but I recommend you to put the weight at 1.
Other styles coming soon.
If you have any problem try copying generation data from my examples, or send me a DM
I can't wait to see what other people make with this :)
Description
FAQ
Comments (20)
Awesome lora and thank you so much for making this lora.
Great. I really like that lora. I just try it and it is beautifull.
Realistic version please :)
working on it :)
You and rolladex be on point lol!! This lora is a mainstay for my ZIT flow. Thx a bunch
im glad you like it :)
Tried your lora, it really gives the amazing look. My only opinion is you can improve is when upscaling or using face detailer, your lora will force a lot of water and other liquids in people faces, even when you prompt to not do it.
That have absolutly no effect for me, i use Yiffimix ckpt, also with trigger word
YiffyMix is a NoobAI model, my LoRAs are for Flux, Pony, illustrious, and ZImage :)
NoobAI is illustrious 0.1 with more data. It works welll on NoobAI alone. YiffyMix uses ancient models. His currenty version uses a new experimental version called Chenkin Noob, Which this should work on....but it doesnt. Chenkin is overall not a great model.
Pretty much any other Illustrious or NoobAI model this will work on. Even Hassaku or Janku has no issue.
@VelvetS The ZiT versions are incredible btw
@LatteLeopard thank you so much :)
lovely, but a question. How to achieve low contrast images like the examples on DarkLinesV2? What is the weight? I can not have same output with same Ksampler values? Using it on WaiIllustrious. thanks
the data is all in the pic, i don't remove anything
Keep in mind that i generate in my own PC using SD Forge not in Civitai's generator
@VelvetS appreciate the feedback, I use SD Comfyui with Wai model, but even with same CFG and sampler, i can not have that dark look. But still great Lora this is , thank you.
Hmm the lora doesn't appear in the stable diffusion web UI if i put it in the correct map. :(
which version? and what checkpoint are you using?
@VelvetS Hi Thank you for responding!
Ehm mind you i'm not too experienced in the technical side of things here but this is what I know.
I'm using a Web UI of stable diffusion. I did some digging into the settings and found these "versions" though i'm not sure it helps:
version: v1.10.1 • python: 3.10.6 • torch: 2.1.2+cu121 • xformers: N/A • gradio: 3.41.2 • checkpoint: a5f58eb1c3
The checkpoint there says "a5f58eb1c3" but i'm using: waiIllustriousSDXL_v160.safetensors or "WAI-illustrious".
I don't think it's the checkpoint's fault because the issue is that the this specific Lora the ZiT Detailed Anime isn't showing up in my list of loras so i can't add the "lora tag". All other Loras you got here do work however! :(
@Sissyleona perfect this is why i had to ask about the checkpoint, you see the ZiT LoRAs are for a model called Zimage Turbo, they do not work with WAI-illustrious-SDXL and do not show up in Web UI Stable Diffusion.
Check out my LoRAs that have the word "illustrious" in the start of their name, like illustrious Dark Lines or illustrious Magical Lines, etc
These ones will show up in your Stable Diffusion
ZiT stuff show up only in ComfyUI and are only for Zimage Turbo Checkpoints
Let me know if you need anymore help :)
@VelvetS Aaaah!! I see. Normally I notice that but i've looked over it. XD i'm sorry. I guess i was used to your illustrious loras. haha. Thank you for clarifying!