saphirenishi

top with knot

i think the consistency with the face is finally getting there. pretty happy with this one

monarch ai lora

for full transparency, this is a text to image AI render and i photoshopped the hell out of it

the hair ornament doest really wanna come out the way i want it to. not really sure why that is. i think the whole thing looks amazing though

for anyone that didnt know, this is the original monarch character im using in my 3d renders

face - light control

bit more testing with how much control we can get with lighting when using AI. and the more im using it, the more im completely convinced we can have full control over what lighting we will get. i can go from a simple 3 way light setup (portrait lighting), to a ring light, to a sun beam light, to a complicated multi coloured cross beam interior light setup with direct and indirect lighting (what we see in a dancing/disco or fashion shows and the like).

what i was afraid of what would happen is that the randomness of the AI would kick in on the face, but it actually works really really well. i made a gif so i have a sequences of the face to make it easy if things change or not.

AI is not your work

ok, so lets play a game.

this is jill. jill is a character that is very popular right now with the DAZ community. people get lots of views cause she is a known and well liked character. so makes sense to use her. these are all 3d renders done in daz.

now my question is, how many different artists did i use up here to create this collages. is it 1? is it 8? or is it something in between?

now when people talk about AI, they say, "this is not your work. if i use the same prompt as you i get the same results". which i agree with. but i dont see a problem there, which other people seem to have. cause to me, i know no one can tell me how many different artists there are up here. these are all the same, all daz renders, all using the nvidia render engine, all using the same character. yet this is somehow not a problem, but using AI is. and the more im having this discussions with people the more i get to the point, that people dont really have a good understanding of what digital art exactly entails. even when they do it themselves.

now i could be completely wrong, but the way i see it, digital art, is a way of expressing yourself by using external software to create an image (or a 3d print or an animation). we not drawing lines, like traditional artists do. we use software to draw that line. wither that is a 3d render, photoshop, or now AI. and with that, cause we use software, the line that i created, can be created by you as well, as long as you know how i set it up and which settings i used. so that goes for a simple line made in photoshop, but also for a fully complete AI generated image. or for using a (smart)filter inside adobe software, or setting up a gloom effect inside a daz render. we will always have the same results as long as we use the same settings and same content. yet, somehow people find it all wrong when they talk about AI and find it perfectly fine everywhere else, which is really weird to me, cause isnt all the same thing?

now please, im not calling out anyone in particular. i just think this is a discussion we really need to have, cause together we decide what the future looks like, so lets make it a good one together!

light control

and an other experiment. here im trying to have full control of the lighting. stable diffusion has this new thing where we can control the lighting in an image. which is really powerful. the basic idea is you have an overlay black and white image where you want the main lights to be, and the ai fills that in for you. its kinda like placing light sources in a 3d scene. which again removes a lot of randomness when using AI 

fancy black lace evening dress

playing with lace patterns, i thought this would give a lot of problems for the AI, but it works surprisingly well. its a perfect symmetry we got going on

open long jacket

so this is that nobody character again. im really beginning to like her! modified the dress from the previous picture a bit, and added a long open long jacket, again all done in in marvelous designer

this combination of marvelous designer and AI is so powerful. within a couple of hours, i have a complete new outfit, and a render with said outfit that looks so amazing. we living in the future now

dress consistency

and an other test. here im trying to see if i run these render through the ai, if it will keep the dress the same across all images. since consistency still is really tricky to get with AI. seems to work kinda nice, although its not perfect

i guess its pretty clear, but the top row are daz renders and the bottom row is what the ai is pooping out

i havent fixed the faces and fingers in these test renders, since they are just that, simple tests

capt america - daz

daz version of the capt america outfit. works great as long as im not having any cleavage. i really need to figure out how to get better results when we using the cleavage slider

only thing left to do now is setting up the texturing

two layered autaum dress with hollow chest

so this is that nobody character from the last post, and wearing a dress done really fast in marvelous designer. i so love how this looks :D

face consistency

and an other little experiment. trying out how consistent the face is if we trying to render a specific character.

the first one i tried was with the always beautiful strahovski. which seems to work really well. its using the lora networking system, so it can be use on top of any other dataset, which makes it really powerful to use, though changing things like hair color or eye color can be a real pain to achieve, since its really sticking to the images it learned from

next one is a bit more interesting, since its from the nobody series. which means its not based on a known/real person/character. there are a couple of these now, but what would be really interesting is if would could create something like this ourselves with our own OC. no idea how one would set a this up, but im definitely looking into this. this is a textual inversion, so its purely reliable on the text prompts you put in, so thats makes it very versatile but not very precis when it comes to smaller details on the face.

duke - O-bikini

as requested

monarch - anime

as requested

capt america 01

capt america ai version

this is why i think we going to use AI for rendering. so the left is the clean render. with texturing that is far from done, the right is the ai version of the left picture. if you look at the ai version there is no need anymore for normal maps. you can tell the ai to make things asymmetric, get a lot better results on the star right away. and all that in a few minutes. while setting everything up in DAZ to get the result we getting with the AI is going to take me 2 days.

capt america dress 02

md work is all done, now for the hard part and start making it all work inside DAZ.

i can already see a huge problem with the star on the chest when changing boob size. also cause of the zero shape, the waist and shoulder look kind flat. making the waist smaller (like i did here on the examples) and putting some muscles on the shoulder will make it all look just a lot better

wasteland - grunt

wasteland - rogue

capt america - dress

first setup

wasteland - wanderer

concepts

marvel dresses concepts

wasteland beauty

flared business outfit

flared business outfit (g8f)

tank top blonde

tank top

more nudity

just pushing it a bit further

red head on bed

cleavage

now thats a cleavage you wont get out of daz!!!

selfie

selfie

big mf

fluffy beard!!!!

casual dress - blue 01

an other take on the previous post