I’m wondering, are simple users able to use the AI?
If so, how?
Can we use it to create lifelike characters and creatures?
Are we able to give it a diagram and image and it produce a character?
Can it make anatomies of creatures? Such as for hunting games?
etc
Hi, Although 3D AI tools have come a long way they aren’t quite at the level you are looking for yet. Image generation tools for 2D images are quite photoreal at this point but AI-generated 3D models are still in the early stages.
Here is an example of an AI model generator for animals available inside Omnivers today, as you see it’s a great example of where we are headed but not quite “game ready”. Omniverse Animal Explorer
For characters, AI is being used to drive animation and provide interactive discussions as well as generate voices but there aren’t many tools yet to generate actual meshes. You can however use capture tools to generate meshes and animate them using AI.
There are interesting tools that use text based AI to generate characters: Tafi - AI text to 3D preview
There are great tools to generate custom characters that can be imported into Omniverse like Reallusion iClone and Character Creator
We just did a great live stream on character creation and animation which you can watch here: Designing 3D Characters With OpenUSD and Reallusion | Omniverse Live - YouTube
We also have a great session on bringing characters to life using AI coming soon: Bringing NPC Characters to Life With Generative AI and NVIDIA ACE for Games | Community Stream - YouTube
The great news is that any interesting tools you find out there to generate 3D models with AI will most likely be able to export to Omniverse where you can assemble and extend toolsets to your needs. Omniverse is a great platform to help build pipelines connecting 3rd party tools and enhancing it with your own extensions.
I suppose what I’m looking for doesn’t exist yet. I’d love to know when it does.
The AI would need to be able to take various 2D anatomy … blueprints? Charts? Whatever the word is… Picture? Whatever…
It would take those and scans of what the creature looks like and it would make a creature with anatomy.
Another tool, like a slider, with X, Y, Z, and rotation based, would allow to see the internals of the creature without damaging the mesh. Slide it to hide part of the creature, then slide it back to reveal the creature.
I’ve seen some models like that but I guess those took at least a few weeks to make by hand. I wish there was an AI that could take those inputs (anatomies, physiologies, looks) and make creatures from them.
That’d be good for hunting games at least. Better than current ones anyways. More realistic.
And something for way later, interactiveness between meshes so they aren’t ignoring each other and flopping all over the place, for things like Fighting games. Make collisions more realistic. What a dream.