Dear Pilou, I followed your instructions to try to test and train a model with Leonardo AI. I also signed up for a month so that I would have more possibilities.
However, the result I got with my pre-raphaellite girls was very disappointing compared to yours.!
I have also the impression that many of your portraits have then received post-processing afterwards (floral insertion, front and back of the ubject etc.).
I also encountered some difficulties with your nice suggestions of 25 May on this blog: with a found image on the Net prompt "post appocalyptic devastated city".
In this case I wasted dozens of tokens, because most templates simply leave everything blank.
The only one that started working was the Dreamshaper v. 7 model , and only with: Controlnet activated and only if selected on: depth or image, otherwise it produces nothing.
And also I got different designs from yours that you obviously used a different template....
I believe that your valuable blogs, and your fantastic inventions would be more appreciated if you also indicated some of the steps you encounter: type of model, slides, promt etc. I think thet a fuller step by step would be much appreciated by all your loyal followers!
<<
Another small question. How do you get portraits in which several studies appear? Even 10 faces in different angles.
Does something have to be written in the prompt?
The different angle are given by the different original images taken!
I never write a prompt except the 3 ones ...
"Portrait"
"Color Portait"
or last day
"Color Portrait smiling"
for the number day 24 as you can see it 3 posts above!
one of the result! :) (i will try to refind the original)
Give me an image and see the result i obtain... with my Preraph data set!
no more than these 5 images!!! My "Preraph" data set!!!
I believe that was this one! :)
here 0.17 so maybe with a more intensity 0.19
You have never the same result if you don't use the seed number!
As you can see here some post process must be made... :)
I don't make it i just will generate an another image! :)