LORA trained on 33 publicly available images of Hailee Steinfeld.
Trained on base SD 1.5, should work with most models (works best with serenity v2).
Use epicNegative in negative prompt otherwise the forehead is very large (probably because I increased training steps)
Use ADetailer if details are lost and strength of 0.8 should work on most prompts.
Consider leaving a review if you like the model. All feedback is appreciated.
Description
FAQ
Comments (8)
this new training method is awesome, better result on low weight plus small file size, hope you revisit your old loras with this new method. :)
That would be a lot of work, since I don't have the training data anymore.
Time to buy a bigger hard drive ig
@manikb any tips or places you learned to make your lora's?
@manikb thats fair, maybe your recent ones like
Selena Gomez
Chloe Grace Moretz
Anya Taylor Joy
:)
@vipersick1000231 Mind-Blowing Deepfake Tutorial: Turn Anyone into Your Favorite Movie Star! PC & Google Colab - roop (youtube.com)
This channel has really detailed guides.
But the main thing is to experiment with different settings to find the one that suits your setup the best.
Watch and learn from these videos, then try your own thing once you understand what each of the parameters/settings do.
@wotelif169768 is there something specific the new models are better at compared to the older ones?
The older models work really well with lcm and low steps compared to the new ones.
Other than that, haven't seen a massive difference between the two.
@manikb the new loras only need .6 or less to apply its style, while the old loras need atleast .7 or for some like selena gomez .9, I like using loras like how i use salt when cooking, if I can achieve the same effect with less of it the better. :)
Awesome👌👌👌