- This topic has 9 replies, 6 voices, and was last updated 2 years ago by
.
-
Posts
-
...he keeps dropping things.....
linkRan it through img2img, hope you don't mind. Some subtle changes but basically the same. Lighting a little better and the people look a little more realistic. With img2img I probably didn't need to use a prompt at all but I believe it helps.
[Photorealistic render of two women in a modern office, one with blonde hair and a large buttocks in a short blue skirt bending over a desk, the other in a business suit sitting in a chair, other male office workers in the background, detailed office environment, 8k resolution, cinematic lighting, unreal engine]
I like how its mostly the same. Not sure why it changed the other woman.
May have to see what it would do to my other renders.The CFG value/denoising strength in img2img determines how much the image changes and how close the image will be to your prompt. A low setting will not change the image much and a high setting will generate an almost completely new/different image from the original. So perhaps it's advisable to use a prompt in img2img to give it some extra guidance.
Please god save us from those humongous KardASSion arses. If they live to 70 they'll need an arse-girdle to stop it dragging along the ground.
Looks very nice, would be interesting when AI becomes even better to create scenes in DAZ for Ren`py games this way. And then with AI make the characters and scenes more real.
If the poses could stay the same as in the original that would be perfect.
Repeatability for continuity would have to get a lot better for comic storytelling.
Eg. the change of the seated character above.
There has to be an effective way of locking certain features in.
Not only Poses for example but also Expressions would be important, I don`t think AI can effectively be doing this with the quality of models that is created so far.
Most of them had problems creating good hands for a very long time..
- You must be logged in to reply to this topic.
