Place in Negative Prompt. Can weight as high as :3 but I would stay around :1 - :1.6As it is trained on flaws.example :Prompt: 1girl, twintails, lipstick, glasses, smirk, medium breasts, slender, boots,Negative Prompt: low quality, &lt;lora:Negative_LoraV1.13-000005:1&gt;Yes you can negative weight in the positive prompt but it changes the whole image if you go to far.<h3 id="this-is-an-experimental-tool-when-you-have-a-pic-you-like-but-has-minor-flaws.-you-use-same-seed-and-regen.-i-use-it-from-the-start-in-negative-prompt-and-go-all-willy-nilly.-klmaz6caf">This is an experimental tool, when you have a pic you like but has minor flaws. You use same seed and regen. I use it from the start in negative prompt and go all willy nilly.</h3>To each their own.=-----------------------------------------------------------------------------------------------------=IMPORTANT FROM GENIE_AIPLEASE READ MSG FROM THE CREATOR OF THE GENIE SERIES MODELSTo help me save up for new hardware to really blow you out of the water, send whatever you wish to <a target="_blank" rel="ugc" href="http://paypal.me/puzzledust">paypal.me/puzzledust</a> . I've got more than enough buzz for site training thanks to many generous community members, however the data set I have been preparing for the next model is over 900,000 images and CIVITAI can only do 1000 at a time. So it comes down to training locally. Its a dream that is reachable but I need your help to get there! There's enough community members who appreciated my creations as much as I do. I love doing it. Seeing how far one can go with all this untrained corners of sdxl and sd1.5. and fill in the gaps, with a deeper understanding of human beings and the smaller detailed usually lost in the noise of mass training. I love sharing them with all of you. I have so many characters and creatively new approaches for training in new ways to handle issues that haunt us all, hands, upside down faces, freakishly unproportionate legs ect. , I have made lora and am testing them for solutions to these, the answer isn't in the LORA's its in a better trained checkpoint. I want to rival PONY and ILLUSTRIOUS. USING TRADIONAL SDXL AND SD15 TAGGING AND a more dynamic approach to image training to better help the checkpoint better hone in on hands and things in the noise when generating images. I am choosing to go that route so others can still merge their checkpoints they put all their time into and train on with no changes to their workflows or VAE. So please. Spare some change. Send a mouse to college. Tiny Timmy Needs a new computey.. I don't really need a new computey for me. I need a new Putey for the community !!! <h3 id="whoever-donates-the-most-gets-free-commissions-for-life!!-ojfja16e6"> Whoever Donates the most gets free Commissions for life!!</h3><a target="_blank" rel="ugc" href="christopheroettle@gmail.com">Msg me here about hardware itself</a> <a target="_blank" rel="ugc" href="http://paypal.me/puzzledust">Donate Here to contribute to our next big models</a>you got SLOW GENERATIONS ??? READ THIS article I wrote immediately ! <a target="_blank" rel="ugc" href="https://civarchive.com/articles/10852">https://civarchive.com/articles/10852</a><a target="_blank" rel="ugc" href="https://t.me/+bgr141jTbO8wNmIx">https://t.me/+bgr141jTbO8wNmIx</a> GENIE_AI Chat Room and Lora Sharing on Telegram

Files with SHA256: bca51914f57e62af0abaef1d8a284a40bd88095b8a4b1d626b28c2c6fb3d6047