Krea2 update <ol><li>Trained on site with Krea2 base. Example image generated with Krea2 turbo fp8. </li><li>The trigger word is “taupriestess” no cap no space. </li><li>The default outfit is a white cape-like dress. If you want the see through outfit of the original image use the prompt “wearing a wide see through sleeve”. Also it helps to hire the lora strength to 1. </li><li>Prompting for hooves is not necessary but when the lora strength is low it helps to keep constant results. </li><li>Putting the strength of the lora to 1 will result in more style bias. If you are aiming for a different style lower strength to 0.8 or lower. </li><li>Difference between epoch10 - epoch9 seem negligible but because of vague gut feeling I have uploaded both. </li></ol>V2 update<ol><li>There was a text bubble image in the data set which I suspect caused Anima to generate faulty text output. This was bothering me greatly so I cleared these images out of the data set. </li><li>Try to get rid of the red artifact the priestess is holding on the left hand. Key word try.</li><li>The lora in V1 seems to be over trained on white background making the result too bright. I added some different backgrounds to mitigate this error.</li><li>Add more different sideview, top down view and more different expressions of the priestess in the data set. </li></ol>Sorry for the inconvenience. Problem number 1 was really bothering me so I wanted to address it ASAP.This lora is based on a single image drawn and designed by the artist Mossanannibalis of a Tau character based in the Warhammer universe.<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/bd3214f9-43b1-4f46-afbd-9789d08e2487/original=true/bd3214f9-43b1-4f46-afbd-9789d08e2487.jpeg" />This was the single image of the Tau priestess.I made this lora because I like the character but also I knew the character was so niche that nobody will make a lora out of this character. So as the saying goes ‘the thirsty digs the well’ I made it myself.I created the data set using chatgpt so the original moss artstyle might be fused with chatgpt style. So with this lora you have to expect moss's style fused with Chatgpt style. I am personally ok with this. There are slight errors in the data set. Especially the character holding something on the left hand. I expected it have tainted the lora but I am at peace with it. Also I gave up on four fingers, it is too much effort for little gain. In my head cannon Tau has five fingers.  The trigger word is “@taupriestess” no space. You can add addition tag to most the image such as, blue skin/ Grey skin, yellow eyes/ orange eyes, these two descriptions can be changed depending on what model you are using.

Mossacannibalis Tau priestess

Krea2 update <ol><li>Trained on site with Krea2 base. Example image generated with Krea2 turbo fp8. </li><li>The trigger word is “taupriestess” no cap no space. </li><li>The default outfit is a white cape-like dress. If you want the see through outfit of the original image use the prompt “wearing a wide see through sleeve”. Also it helps to hire the lora strength to 1.</li><li>Prompting for hooves is not necessary but when the lora strength is low it helps to keep constant results. </li><li>Putting the strength of the lora to 1 will result in more style bias. If you are aiming for a different style lower strength to 0.8 or lower.</li><li>Difference between epoch10 - epoch9 seem negligible but because of vague gut feeling I have uploaded both. </li></ol>

Files with SHA256: 01fcaec03c43dac4bb6fc7f485b71bdfc0c57ec16ccdd1ae0a2460644dd146e8