Home Artists Posts Import Register

Content

trying to figure out how to do better eyes, for now do one, then photoshop the other one.

as for hair, im not doing 3d hair anymore, all ai now. so much better!

Files

Comments

Dolomaticus

Well that leads to questions about how well can it do hair. We should talk on that since it could finally get Red rounded out. Also is this AI program better than the first one you was using, or just different?

saphirenishi

its different, its because these ai's use different datasets that contain different sets of images they work with. in the near future we can actually start adding our own images to these data sets as well. its actually already possible but i do not understand the process to make that all possible. but check this out: https://www.youtube.com/watch?v=W4Mcuh38wyM&t=330s&ab_channel=CorridorCrew if we can do this too, then the possibilities become endless. we can add red to the the data set, and real life hair do's, and then let the ai mix and match the whole thing. only things that not seems possible is nudity. but we are smarter then an ai so we can make that work!

stoperArt

One think I like is the details it adds to the hair. On the second example it went too far, but in the first one it kept your original styling and posing and just added detail. The question is, would it maintain this in a series of images? If you rotate the character a little, change expression, etc.. would the final results look like the same character with the same hair?

saphirenishi

yeah, thats exactly what i was thinking about doing hair. hair is a pain in 3d, so if you can do it easier and better with ai, why not use that. in the second pic, its not that the ai went to far, that was me pushing it to far. i could have go way closer to the original hair piece, but im just experimenting how far we can actually push this. and the freedom we have is actually kind of amazing now consistency is a problem, as long as the ai doesnt know a specific character, its going to be really hard to get the same face over and over again. so we need to be able to add our own pictures to the dataset, and let the ai know who saphy is. then in the prompt we can let the ai know we want saphy as the character on a picture, and is will poop out saphy every single time adding our own pictures to a dataset is already possible, and i have looked into it, but its a very complicated process for now. so ill just wait till they have an easier way to make this all happen