Hello Pilou,
congratulations on your masterpieces and on the very useful blog full of excellent and fascinating ideas. I wanted to ask you several things starting with the first one. What exact promp do you use with Leonardo to get these beautiful Pre-Raphaelite female faces ? with these lights and shades ? And also what promt do you use to get your beautiful surreal engines ?
Especially the image in the blog number: 2141, 2144, 2154 e for the engines 2155.
In the next few days I will ask you about the copyright of the images obtained with AI, whether all programmes have the same limitations or whether for example programmes such as Stble Diffusion do not create problems for the use of the images.
Thank you
Nice nick poilu, I am an historian by profession and I appreciate it!!!
I use no Prompt! Just only "Portrait" :)
Load any image and launch the Training Model engine! :)
I use a personal training engine by Leonardo made with only 5 images!
As a free user i am afraid that you can't any more make that! :(
I was on the start of Leonardo so I have the luck to make 2 training engines Model! :)
I made all errors of beginner... no size of the training image (512, 512), different styles etc...
but anyway all works like a charm!
Take any image and use your model trained, no prompt ..it's a delight for the eyes!
As you can see i must add 2 black sides panel on each images for have a square of 512x512 ! :)
(size of the training model)
I have no more than that! :)
the trick is to load an image and play with the Init Strength of the "Image to Image"
0.07 is the big deviation ... 0.6 is near the original
i use 0.17 and 0.27 for have a difference from the original
It's magic with only 5 image you can make anything! !
That works even with anything ...landscape robot even it's not the same subject ...
Sorry i have no time and Internet now more so wait on big week for more details my return :)
About the lights nothing again all depends of the original image!
Generally its impossible to recognize the original image when a small Init strength is used! :)
About the prompts of "engines" : it just "complex engine"! if i want some landscapes that will be something like
"complex engine on the beach" or "complex engine at mountain" etc...
About copyright if there are watermarks... i suppose there is no problem...(if there is no sale)
About the lights of engine...it's depend of the "render AI used"
some have natively different styles like Leonardo, ClipDrop...
other like Bing have not so you must put something inside the prompt!
The other solution is "Image to image" of an image imported... so light will be the same family of the imported image...
you can add of course other thing inside the prompt...
For the engines images you can input an engine from another IA inside an another and make variations! :)
Like here it's an only one Bing "engine" model inside Leonardo for variations ! It's an infinite game! :)
but impossible to know it's a Bing Model if i don't say it! :)
Big Zoom: https://imgbox.com/g/XA6KKVLEzI
Great Pilou, you were very clear, as soon as possible, maybe when you are more free, I will bother you again. Do you also have an email possibly to contact you? My grandmother was French, and my nickname also refers to the French painter Fernand Leger, I can also write to you in French, I'm Italian.
