Download

Meeting Quality Bar of Kite Demo

Why are there no assets on the marketplace yet that meet the visual realism of the kite demo?

There are a thousand PBR tutorials, including a breakdown of what was done with Kite, but I keep banging my head up against the elephant in the room.

I’ve tried so many times now to reach the quality bar but keep falling short. Is there some special undocumented recipe for success?

Cheers :slight_smile:

Normally it’s just about your 3d art + fx skills :slight_smile:
I personally think that the reason, why you cant find any assets in such a high quality on the marketplace, is, that the kite demo assets are not directly made for a game (high poly, high res textures,…) but the ones on the marketplace are for games.

Epic went to an “e}{treme” detail to show off their Engine.

Not everyone is willing to go that extra mile to make sure the leaf on that tree looks as real as it can be.

So we’re simply not at the point quite yet where rendering like that can be applied to traditional gameplay?
Makes sense I guess.

I think we are, it’s just that you don’t have the audience for it yet. The computing power is there, but not in the hands of the “many”. Your multi-gig project will crush the machines of the masses, thus relegating your project to obscurity before its time.

Did you see the actual effort they went into making the assets for it? Including flying out to Scotland and taking a few thousand photos + lightprobe photo’s with some reverse calculations that was purposely written just for this then setting up a custom pipeline to create the objects from photogrametry. No one but a full studio (or some serious art folks :P) would have the ability to get this quality of assets. And that was the point, to go over the top to try and produce the best quality assets you can get in today’s engines.

It’s like buying a Ford then looking at a Ferrari and saying - but I have four wheels, a steering wheel and an engine … why doesn’t my car go as fast?

Of course it’s possible, but as Yggdrasil mentioned, the audience for that is missing -> not everybody wants to spend over 1000€ for a good PC that can run such high quality games. Of course this will change in some years :smiley:

photogrammetry you then have evrthing it is like a picture so thats why

There are actually some rock pack on the marketplace which were created using Photogrammetry and I’d say those are up to Kite Demo rocks quality somehow. You can’t get the same level of detail on something by hand (or it takes a huge amount of time people don’t usually want to spend).

The big thing with Photogrammetry are the details. You can sculpt and model high poly, but you can’t ever get perfect recreations of details, since the level of detail in the real world goes down farther than subatomic particles. So, with photogrammetry, you are measuring that detail with light, and recreating it. Now, you can use a low poly object and place those details on them, and will get more realistic looking detail than spending 10 times as much time detailing it in Zbrush.

But Photogrammetry is more expensive than Zbrush, if you want the best quality, and is a memory limited operation on a computer, meaning a computer with 128GB or memory will doe the job faster than a computer that is twice as fast with half the ram. Multiple GPUs help, as long as the program you use uses them. Agisoft is great because of this.

You don’t have to go to exotic locations, since the thing you want is detail, and colors are secondary to this. So if you want foilage, take an interest in some scientific plant classifications, find cousins of the plant you actually want the details for, and then manipulate the colors at the end to look like the desired plant. Bricks are the same around the world, and rocks can be found at your local swimming hole. Bark on trees are varied, but you can find endless assortments all around the world.

All of this takes time to shoot tons of pictures, and then the computer processing time. In the end, they had a team of more than 20 people over 3 months times, which is something like 2160 man hours, not including all the tasks they let run overnight to do the actual photogrammetry. Building a library of these types of assets that can be reused will be worth the time, if you have the time/money to spend to do it.