Anonymous at Sat, 25 Jun 2022 21:09:41 UTC No. 905249

>>905246
Questions threads exist. Use them next time.
Photogrammetry exists. Look it up next time.
Welcome to /3/.

Just let this thread die everyone.

Anonymous at Sat, 25 Jun 2022 21:30:34 UTC No. 905253

>>905249
>Photogrammetry

then why bother making 3d models by hand? You will never reach the quality of the real stuff !?

Anonymous at Sat, 25 Jun 2022 21:49:06 UTC No. 905257

>>905253
Because photogrammetry is still pretty shitty, and the end result still needs to be cleaned up by hand.
Not to mention, you can't do photogrammetry on something that doesn't exist.
It's not a shortcut to good work, it's just a tool like anything else. You get out what you put in. And to get a good model, you still have to put in work.
If you know what you're doing, it's often cleaner and quicker to model it yourself.

Anonymous at Sat, 25 Jun 2022 22:06:40 UTC No. 905261

>>905257
Lets say a team wants to make a medieval combat game. Could they just record a video of people in armour fighting with swords etc and put it into the game? They wouldnt have to model the weapons, armour, men, animations, etc, everything is taken from real life, is this possible? If so, why arent the game developer doing it this way ?

Anonymous at Sun, 26 Jun 2022 00:23:41 UTC No. 905286

>>905261
Do it yourself and get back with us.

Anonymous at Sun, 26 Jun 2022 01:13:46 UTC No. 905305

>>905261
Let's break it down, shall we?
>Could they just record a video of people...
Video-based motion capture does exist, where you can take a video from one (or preferably more) angles and use AI to do pose estimation (note the ESTIMATION here). But the result will be as bad or worse than a dedicated motion capture setup. Even direct mo-cap with state of the art tech still needs cleanup. It's drastically less, but you can use your head to compare that with a jury-rigged solution with a single camera and no other data. You're gonna be doing a lot of cleanup of the data, and for some parts you might even have to animate yourself.

>They wouldnt have to model the weapons...
Scans are used from time to time in game, but more as a baseline reference. To be usable in a game (ignoring the new shit in UE5), it still needs decent geometry and still have a justifiably small polycount. You can't just throw a bunch of scans into a game and expect it to be performant. Generally, an artist will take a scan, retopologize it (essentially tracing over the model, making sure to have good edge flow and geometry), and then project the textures onto the model from the scan.
We're only talking about static objects here too, not people. People are another can of worms entirely.

>If so, why arent...
In the case of motion capture, as mentioned previously, they have better setups than a single camera to do pose estimation. They simply don't need to.
As for 3d scans. Scans by themselves are messy and have millions of points floating all over the place. It takes time and work to get them to a level that's usable for anything more than just looking at. Time and work costs money, and for reasons that should be obvious, that time and work would be better spent modelling something from scratch than trying to clean up a bunch of scans.
Scans have their place in a workflow, but at the start of the process, not the end of it.

Anonymous at Sun, 26 Jun 2022 01:21:05 UTC No. 905308

>>905261
It's a lot cheaper to hire an artist to model the thing you're after than obtain real world artifacts,
fly people there, get permissions, bring the gear and do the scanning process + clean-up work.

The fact that you need access to something rare and expensive in order to 'simply scan it' kinda makes it way less useful than having the ability to synthesize anything from the ground up.

Anonymous at Sun, 26 Jun 2022 09:55:37 UTC No. 905372

>>905305
>>905308
Why isn't it possible to scan an area like google map cars and then digitalize it/turning it right into a video game environment?

When do you think this will be possible? You could create the most realistic environments, forests, cities, indoors. The software should also recognize the stuff and automatically give the items their characteristics, like bottles on a table, chairs, cars on the street, people...... A powerful AI is needed for this i assume ?

Anonymous at Sun, 26 Jun 2022 10:59:20 UTC No. 905386

>>905372
>Why isn't it possible to scan an area like google map cars and then digitalize it/turning it right into a video game environment?
I already told you, because it takes work to actually make it usable. What a scan shits out is just a bunch of randomly placed points that make up an object. What you need for games is a bunch of points that are nice, neat, orderly and most of all, efficient. You ideally want to make an object with the least amount of points (vertices) as possible. Scans are literally the exact opposite of that.
That being said, tech like that is in use already, but less so in a production environment.

>When do you think this will be possible?
Now, and a bit further on from now. AI generation can already create decently detailed models from photos, but they're still not so good at creating those models with proper geometry. Something like that is a little ways off, and would probably need a bit of ingenuity to accomplish. Since there's not "one solution" for any given model when retopologizing, and everyone does it different. One dude might retopo a statue one way, and another dude a different way, but both are equally as good.

>The software should also recognize the stuff and automatically give the items their characteristics, like bottles on a table, chairs, cars on the street, people
At this point you're just talking population algorithms and shit. That's all standard fare these days. Not so much with an AI, but you could easily add a tag to an asset like say "table" and have it automatically add clutter to it or other assets. Plenty of games do this now. Not many studios place every tree or every instance of grass these days.

AI is a good way off from being a complete shortcut. Stop trying to use it as such and start modelling.

Anonymous at Sun, 26 Jun 2022 13:21:43 UTC No. 905405

>>905386
>and start modelling.

No, modelling is replication of real stuff, its like trying to paint a scenery as realistic as possible altough it will never be as real as a photograph.
Its tryhardism.

Anonymous at Sun, 26 Jun 2022 21:20:02 UTC No. 905474

>>905405
No, modelling is not just a replication of real stuff. Don't be retarded.
Are spaceships real? Aliens? Dragons or anything else someone can come up with in their imagination? Good luck scanning all that nonexistent shit.
If your only gauge is realism, you sound extremely ignorant of the possibilities. Sounds like you've just got no creativity at all. Anything you can imagine, you can model, but anything you can imagine does not exist in reality.
Mimesis is not the end goal of 3d or 2d.

Anonymous at Wed, 29 Jun 2022 18:33:43 UTC No. 905968

>arguing with an ESL
>arguing with a knowitall
>arguing with a nomodeler
You guys are wasting your time with OP, he's just fucking lonely or something.