Keep in mind this is not even in Beta form and the results are no where near complete to final form and it can only do single objects for now but in a year you will see full reverse engineering of interiors and exteriors from a couple sample images
1 Like
Well i think ,
its a possible future… and will be a cool tool to artist too,
For now its just a demo, you can clearly see the difference in the trumpet from the original 3 cilinder in the middle and the almost flat 3D you can see at 34seconds.
But even if it will be updated in a perfect model in a couple of year
cant replace an artist.
The big- cool part of the 3D is making things that does not exist already.
what If you have to model a complex steampunk car, ? a winged octopus ?
If you think to how hard is sometime to set a machine or a software to do what you want,
we have AI, deep learning machines, renderfarms with swarm rendering BUT
we still have also human callcenter XD and administrative paperwork madness XD
i take a lot of fun in modeling and while doing it i often change the project
to try something that the shapes and the spaced comunicate to me.
I draw sketches both 2D and 3D , merge ideas with memory and reference
and also evaluate what i am doing with all i have seen,
and sure, modelling modular or keeping in mind the uv or the skinning deformations
or project limitation … and who know what else…
Beside its a cool tool !
hope it get better like any other tool ,
he can help us making simple stuff from photo? cool,
but replace artists its very likely a very very far thing…
1 Like
15 years ago, I saw tools at I/ITSEC (a military training conference) that took a bunch of photographs, and turned them into a level for some game engine. “reality capture” is not new.
What you will find, is that games aren’t reality. Try to put a first person shooter in a realistically proportioned room, and players will feel cramped, and get stuck on the very narrow doors, and benches and stools and chairs and tables and lamps will all get in the way.
The art of making games, includes the art of making things that aren’t realistic but feel real.
Sure, with enough training, models can probably figure out how to do that, too, at least for some particular art style. And then you need to train the model for your particular art style. At that point, the artistic work in producing a game ends up being more of “this is the style I want” and “these are the characters an environments I want” and less of the pushing polygons around – and maybe that’s actually great! Most artists start their career by making beveled boxes in Maya, but it doesn’t have to be that way. That’s simple work that robots would probably do better anyway 
1 Like
Even if tech like this replaces 3d modelers and texture artists, you’ll still need artists for other tasks like level design and art direction. And you’ll still need artists for models that don’t exist in reality or can’t be scanned.
One big flaw in this demo and the models it produces, the surfaces are very bumpy for models that ideally need to be very smooth and clean. The reflections don’t look as good as they should.