zit_v1.0:
This is the first ZIT LoRa I am publishing. It is trained with the same dataset and same danbooru tag style captioning as the illustrious version.
It can be used with tag style prompting, but I noticed that natural language prompts or mixed prompts seem to work better with Zimage Turbo. So using "detailed drawing in kyhustyle of" or "digital art in kyhustyle of" at the beginning of your prompt followed by your main subject should be used instead of just the trigger word "kyhustyle".
I recommend a weight of 0.5-1.0, but at values close to 1.0 the bias towards korra gets pretty strong, since she is in most of the images in the dataset.
For the showcase pics I used an fp8 quantized version of base Zimage Turbo, so they look a bit different from the illustrious showcase pics using a finetuned checkpoint.

trigger words: kyhustyle, korra

Feel free to comment, if you have something to say about the lora and share your images on the model page.

If you like the lora please have a look at the showcase pics and leave a reaction.

__________________________________________________________________________________________________

ill_v1.0:

This style lora is trained on images by KYHU, the old pen name of artist IAHFY. The images used in training depict Korra, Asami Sato and a few other characters with plain background. So this lora is best used for generating characters without detailed backgrounds. I noticed during testing that the original style has more impact, when omitting the quality tags I normally use for the checkpoint. I guess the reason is the style is a bit rougher than your average style. Works good with tags like "monochrome", "sketch" or "sepia".

trigger word: kyhustyle

additional style tags: monochrome, sepia, lineart, sketch

character tags: asami sato, korra

Feel free to comment, if you have something to say about the lora and share your images, if you generated something good. If you like the lora please have a look at the showcase pics and leave a reaction.

Search AI Models & Addons

very few results