This is a model trained on artwork from DaveCheung/VanillaCookies/TehDave. This is the first lora I've trained on Krea2 and it captured the style well. I didn't use artificial data this time, hoping that Krea2 would be savvy enough to let you prompt around some possible issues introduced by a stricter adherence to the style. I think that was the right choice for now, as it gives a more purist base point to work from. A V2 will probably use artificial data if I make one.

Trigger Word: In the style of vcookiez

The trigger word isn't strictly necessary, but will emphasize the style. Some prompts can weaken the style pulling it towards realism. That's just the result of Krea2's wider training on photorealism. You can raise the weight a little or add style prompts to try and pull it back.

The dataset contains a significant amount of NSFW art and might pull you towards those themes.

If you need to the clothing to cover naughty bits and it's refusing to do so with prompts, try this slider: SFW Filter Slider | Krea2 - V1 | Krea 2 LoRA | Civitai