If you like what I do, leave me a ❤️ to receive notifications of future updates and follow me here on Civitai. Remember that it is a hobby for me and it costs a lot of time and everything is shared entirely for free, I appreciate that you make a review and share your creations on this wall.
Milo Manara Style - By Devildonia V2
This is my third model trying to represent the style of the master Milo Manara. The first one was in Stable Diffusion 1.5 and the second one in Stable Diffusion XL. While both are fun models and achieve some very good images, they are far from the master and only come close to his mastery.
Ah, but... With Flux and the experience gained with those two previous models, things have changed, and we are getting a little closer. This first version of LoRa is really good, but it still has a lot of room for improvement. Things that I have not wanted to implement, images in the dataset that could be better, the training that is experimental...
The dataset of the V1 that I have not published had almost 1000 images and in this V2 it was reduced to 400. Each image was digitally enhanced, rescaled, in other words, a lot of work. But I think it was worth it, or so you will have to confirm.
I have tested this LoRa with several models, both Dev and Schnell, with various resolutions and rescaling, and the truth is that it works surprisingly well with all of them. Obviously it works better with Dev, since it is the model I used to train the LoRa, but with Schnell it works really well.
Sample images were made with Project Gaia V1 (Flux Dev model) and Project HaDeS V1 (Flux Schnell model).
See you in the next update!!
