![user_651468310541608956](https://image.tensorartassets.com/cdn-cgi/image/anim=true,w=40,h=40,q=85/users/common/avatar/651468310541608956/50383225-da78-ac97-b190-337084601722.png)
user_651468310541608956
![Daily Challenge](https://image.tensorartassets.com/cdn-cgi/image/anim=true,plain=false,w=14,q=85/badges/e13137b7-b048-4d79-8c98-e2bb580be0ac.jpeg)
Published:
~ ~ Inverted Eiffel Tower V1 ~ ~ In general, image generating models tend to have problems creating images that go against common sense, (especially if they have not been trained in it) if you ask them to create a man catching a fish they will probably believe it, but if If you ask a man for a fish, you probably won't achieve the desired result. I tried to create an image of the Eiffel Tower upside down using prompts but I wasn't able to (although I didn't try that hard either, but I couldn't think of anything else to try), so I tried to use an image of the Eiffel Tower upside down as a base in Img2Img but the The result did not convince me, so I decided to use controlNET to create the images and although it is far from being a result of the highest quality, it is the method that has convinced me the most. The controlNET that has given me the best results was the Lineart model in its lineart_realistic configuration, the closer to 2 the image is more similar to the controlner model.
Comments