Krea 2 Premium LORAs
Model description
Woah, woah, woah, PREMIUM LORAs ? What's this all about ? I'm glad you asked, since the company that shall not be named decided to call their 100 step abominations "style" LORAs, I had to adapt and somehow distance myself from that, so here we are: Dever's Krea 2 Turbo Premium Style LORAs that were not trained in under 2 minutes to spam HuggingFace. Rant over.
Style Transfer RC1 (experimental)
It was trained to keep the composition of the first reference image and apply the style of the second reference.
Use KSTRANSFER as the trigger. See the workflow example from here
Uses Krea2 edit nodes from Ostris
Count BinFace character
Use Count BinFace as the trigger. Who is Count BinFace ?
Blue Eye Samurai style
Use dvr_bes_style, ... your prompt ...
Cyanide and Happiness style
Use dvr_ch_style, ... your prompt ... as the trigger word, RECOMMENDED to add stick figures to force the stick figures specific to the show. LORA samples show the same prompt and seed with the LORA off and on.
The dataset has a bias towards characters so prompts without any people might not be as great. Similarly very long prompt with a high emphasis on realism might not work as well either.
Blueprint / Wireframe style
Use dvr_wf_style, ... your prompt ... as the trigger word. LORA samples show the same prompt and seed with the LORA off and on.
Prompts used for samples are from Krea's prompting guide
Download models
Download them in the Files & versions tab.
Disclaimer and Important notes
The LoRAs in this repository are created for research purposes only and are not intended for commercial use. Some might be based on TV series which are protected by copyright. Users utilize these models at their own risk. Users are obligated to comply with copyright laws and applicable regulations. The models have been developed for non-commercial purposes, and it is not my intention to infringe on any copyright. I assume no responsibility for any damages or legal consequences arising from the use of these models.