I've discovered that two current models already presented on Civitai are quite poor quality, so decided to train my own, preparing HQ dataset with different angles, emotions etc and was satisfied with the result. Hope you like it!
Description
FAQ
Comments (9)
without any trigger words, this is pretty useless...
Could you please explain the benefits of using trigger words in Lora models?
@seraphinanoire so... I don't know much.... and i've never made one... but I figure that the trigger word directs the AI to use your lora over what it already knows... so if you had 'Aur444d' as a trigger word, it relies on your lora, but for now i get a garbled-up blonde girl standing in front of the aurora borealis...
"portrait of 1woman, aurora, blonde hair, bangs" was a disaster. I ran 40 images, and only one had signs of getting her hair right, so the info is in there, but it's not coming out.
I wish you luck, the images you posted are lovely.
@JoeJo832 Oh, mate, I got you. Do you use Automatic1111 I suppose? Then you should use <lora:aurora:1>, this is the regular syntax for using LoRas in Automatic1111 prompts: <lora:loraname:weight>. If you use ComfyUI, there's LoRa loader node.
nope.. using NMKD's GUI which is easier than Auto1111... I just click a lora to turn it on, and can change the value, which is right next to its name. (it's even easier than comfyui)
maybe it's just not compatible with NMKD, or maybe NMKD relies too heavily on trigger words
@JoeJo832 Sorry than, can't help you with it, because I do not know this interface. I started using SD with A1111, but wanted more, so a bit later switched to ComfyUI, which is so much better and customizible. It looks a bit complicated, but it's only a first impression, there's so much fun in working with nodes!
Can you please do Marzia Bisogning / Marzia Kjellberg (Pewdiepie's wife)
@bififek932725 please this +1
@bififek932725 please this +1