Better than 3D gaming? AI images in 3D.

[costiq]

I know it's a long shot, but does anybody knows an AI image generator that can do real geometric 3D?
LeiaPix Converter does a great job creating fake SBS out of single images, but wouldn't it be fantastic to have the AI do real 3D?

https://youtu.be/v2MkABLlAew

[Losti]

How should it work to create geometric 3D from a 2D image ? This is technically impossible. You need a camera angel shift for this so means 2 pics.

[3DNovice]

....

[costiq]

How should it work to create geometric 3D from a 2D image ? This is technically impossible. You need a camera angel shift for this so means 2 pics.

I'm very bad in explaining things, so no wonder you missunderstood what I meant.
I know you need two angles for a stereoscopic image.

And that's what I was wondering about: is AI intelligent enough to create the same image but from two angles, and render an SBS or top-bottom image?

I've asked Bing image creator to do it (with prompts like SBS stereoscopic, top-bottom, 3D) but, while it tryied, it succesfully failed.

Here's a top-bottom image it came with:






...and here's a SBS one:






If for some people AI seems scarry, it being a cyclops is even more terrifiyng...

[costiq]

If any of the new AI image generators do any type of stereoscopic output, it would likely be applicable to a VR device, so you should watch/look
for VR image generators.

Thanks 3DNovice for the links!
And thanks for the great tip about VR image generators.

[costiq]

3DVision Blog has a manual 2D to 3D photo shop tutorial
https://3dvision-blog.com/2181-converti ... photoshop/

The LeiaPix Converter I mentioned is much more convenient. It does the depth-map method automatically and you can save the rendered image in multiple formats including animated gifs and SBS.




https://convert.leiapix.com/

[costiq]

The LeiaPix Converter I mentioned is much more convenient. It does the depth-map method automatically and you can save the rendered image in multiple formats including animated gifs and SBS.
https://convert.leiapix.com/

BTW, they also sell a 3D tablet: Lume Pad2 (in the US) https://www.leiainc.com/ Nubia Pad 3D (rest of the world) https://ztedevices.com/en-gl/nubia-pad-3d/

[whyme466]

AI-based 3D images are likely to have multiple artifacts. One can get a demonstration of this with in VR (which has 3D display), when NVIDIA’s DLSS is enabled in modded 2D games. I strongly prefer traditional oversampling to DLSS-based upscaling. DLSS, an AI-based technique that NVIDIA has worked on for years, still creates an immersion-breaking shimmering edge around characters and objects with v3, even in Quality mode.

[3DNovice]

....

[treeeddd]

Have you tried this method: Generate 3D short films from 2D AI images!https://www.youtube.com/watch?v=PFybvdKH8hc

AVsforum method for how they convert 2d to 3d video frames: https://www.avsforum.com/threads/a-new- ... 63/page-21

I've played with some 2d to 3d picture conversion with the above methods, I'm limited by vram for some methods. I haven't tried the blender method yet.

[costiq]

Thanks for the links!
I did try the depth map method in Cinema4D.

I just wish somebody with more brain cells then me could teach AI to generate real stereoscopic images. That would be amazing.

[3DNovice]

...

[costiq]

I did see this
https://youtu.be/utBH8ouGQJE WOW! 360 degree AI Images - in Skybox, ControlNET, A1111 & Midjourney // Complete Guide

Many thanks for the link! That's a great find.
Unfortunately, while it does 360 degrees it still can't do stereoscopic.

However, using the LeiaPix Converter, you can get a decent 3D effect in VR.

Here's a very short video I've created: https://youtu.be/1QR-_TVktEc
It doesn't look too bad in a headset.

[3DNovice]

..

[costiq]

.

[3DNovice]

...

[treeeddd]

Hey, nice 3d image conversion. I really liked converting 2d imaged to 3d, but always had some issues with artifacting it caused on the edges, and the inpainting always sucked. I know there's tools out there already for better conversion. But let me what other tools you've come across to make better 3d images. I want to get back into it again once I get my 3d vision rig setup again.

[3DNovice]

...

[3DNovice]

....

[treeeddd]

Stable diffusion is developing at a rapid pace, video animation is looking better, there are better depth map algorithms.

The lume pad 2 can do 3d effect glass free real time for images and videos. Has some sort of stable diffusion app built in too. This would be the best method to view such AI generated images.

But Damn, which the tablets weren't so expensive and wish it could be used as a 3d monitor as well.

[3DNovice]

...

[3DNovice]

....

[3DNovice]

....