The first was a LoRa trained out of curiosity and gave interesting results despite its flaws.
In version II:
-
I changed over 50 images.
-
More variety.
-
I added more images with characters.
-
And closer characters, which I didn't have in the first version.
Various prompts: city ruins, castle ruins, palace ruins, temple ruins, columns, arcades, monolith, tower, dome, arch, covered by vegetation, moss, trees...
Depending on the long-term results there will be a III.
After the first tests the characters seem better proportioned in relation to the buildings.
Strength: 0.7 -> 1
