Saturation is most of the time a bit low, just add "high saturation" to your prompt.
Use "jhyd style" to activate.
Description
FAQ
Comments (19)
large file size! Can it be optimized, perhaps?
I will take a look at it.
I've resized the lora and replaced the original file. The style shouldn't be changed to much. I hope this is small enough for you,
@CowKing perfect, thanks! I'll run it through it's paces and share some images here when I can. Really looking forward to it!
@CowKing the size-optimization may have not gone to plan - I'm not able to get a style from the LoRA at all right now 🙁
@eldritchadam Hmm. I've tried the example images with the resized lora model and these worked with minor changes. Can you please post an image with the meta data?
@CowKing OK - so the issue is apparently that the negative prompt is for some reason absolutely necessary. This is quite the break from typical SDXL behavior. I almost never use any negatives with the XL so ignored yours.
Not ideal, but not a big pain point either and the output is truly beautiful. It'd be nice to see the need for negative prompt mitigated in subsequent versions. It is maybe just an issue of the training not fully converging on the style of your dataset? Train longer or with a higher learning rate?
@eldritchadam I'm happy you like the style in the end. I will look into this negative prompt problem for the 2.0 Version I'm already working on. Also thanks for your time and the positive review.
Can't get it to generate anything, may I have the version from before the update 5 days ago?
I've added the other version now too.
Did you try to use the prompts of my examples?
Also make sure to use the activator text in your positive prompt.
For better style consistency add "photo, render, 3d art" to your negative prompt.
I had the same problem. It turned out that the file for the lora did not fully download. As soon as I downloaded it again and checked that the file size was correct it worked.
@VisChros Oh, this never happend to me on this website. Thanks for this usefull information.
Doesn't work for me at all. The file size is right and I tried the examples here. Nothing.
Hey, buddy, I found a small problem. The file you shared has a ".smaller" suffix, which causes an error when calling lora in SD because "jhyd-step00001000.safetensors" is used. (This is translated by Google, I hope you can understand it)
This model is amazing. But I want to know what jhyd means, ths~
just a trigger word.
"Jihad"
The style on this is unique and iconic, thank you! Flux and Illustrious have seen some great styles, but this one will be the best by far when you release it
style does look really cool, but the model didn't work for me either
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.





