@musashidan we have 2 handheld scanners from this company at work, although not the same model. Each of them cost 20000 Euros + you need a beast laptop with excessive amounts of RAM so you end up spending 25k for a single setup. Maybe this newer version costs a bit less, who knows.
How the scans? Could you build nice normal map details from it? Comparable to photogrammetry like Reality Capture or Agisoft? Does it work for not 100% static subjects (human) ?
I've seen this kind of scanner in action, not sure if this was this exact one, but i guess they all work similar
and ant1fact is right, they are really pricey, i think the cheapest one at the time was around 13000$ without the PC
but they worked really nice and the workflow was better than photogrammetry from what i could see in this 1hour session
the software would calculate the scanned result in real time on your screen,so basicly you see directly what you are scanning, and you could go in and re-scan areas if you didn't like the outcome
you could also scan different areas individual and the software would stitch it together if i understood that correct
i am not entirely sure how good the actual texture quality was, since the example i got was only geometry, but the geometry was pretty nice (car seat). all in all pretty neat but way to expensive if you don't use it commercially
@musashidan we have 2 handheld scanners from this company at work, although not the same model. Each of them cost 20000 Euros + you need a beast laptop with excessive amounts of RAM so you end up spending 25k for a single setup. Maybe this newer version costs a bit less, who knows.
Pretty insane, but like all technology give it a few years and prices should drop dramatically as more competitors enter the market.
I'll answer some of the questions above and include a bit more info that comes to mind. I don't really know the quality of the textures because we only use the geo and yes you can definitely bake normal maps from it after cleaning it up in ZBrush. The quality of the mesh depends on the type of surface / time spent scanning. Metals are literally invisible to the scanner so they need to be coated first. Dark things, and velvets take longer to scan and require more positioning targets (which are also quite pricey but you can reuse them a few times - they're stickers so they last only as long as the glue holds). I think RealityCapture should be able to produce the same quality but the scanners are important from a workflow perspective. We often go to clients to scan items on-site where we have a limited time available, so we need to be sure that we got the scan correctly there and then. Photogrammetry would take longer and would be more prone to errors if we only took the photos on-site and then processed them later. And no, this stuff doesn't work for moving objects, so I would say it's not ideal for scanning humans (although possible if the person is standing still because the scanner mostly cares about the positioning targets' relative position to each other)
Is this hand scanner somehow more advanced than previous models or is the thread just about the class of devices as opposed to camera based setups? I recall we were looking into hand scanners several years ago for doing character work.
Can't remember the name of the device anymore but the price was around 15k USD. It can briefly be seen in action in this segment from UC4 starting at timestamp 2:40:
@thomasp As far as I know the scanned data was primarily used as reference to speed up modelling. The texture resolution and quality was nothing you could simply use. In my experience hand scanners aren't that useful for textured assets. They have a very high (3d)resolution but require a static objects and the texture output (if there is any) lacks compared to high end cameras. They sit a bit between a single-camera photogrammetry workflow and a photogrammetry scan box (many cameras)
Good photogrammetry scans on the other hand are already so good you can use a lot more of the scan result as direct input for the final asset. Though most will post-process them anyway to archive a certain effect or match the assets with the pipeline requirements
But you need a very expansive scan-box with a lot of cameras + software (not that expensive anymore) https://ten24.info/sample-scan/
I love this stuff but it seems we're moving into a direction where assets become more flexible to use instead of just single very high resolution unique textures (at least for games where you always have hardware limits).
@thomasp As far as I know the scanned data was primarily used as reference to speed up modelling. The texture resolution and quality was nothing you could simply use. In my experience hand scanners aren't that useful for textured assets.
Oh we were simply looking at basically doing what is shown in the video I linked:
have a faster, better solution for doing realistic clothes than trying
to painstakingly recreate them in Marvelous. For that we figured all that would be needed is
medium to high resolution geometry, no textures, then straight off to
retopo for sculpting, refining and adapting.
Oh we were simply looking at basically doing what is shown in the video I linked:
have a faster, better solution for doing realistic clothes than trying
to painstakingly recreate them in Marvelous. For that we figured all that would be needed is
medium to high resolution geometry, no textures, then straight off to
retopo for sculpting, refining and adapting.
@thomasp wait.. I get a deja vu! Haven't I red this some time ago - just slightly different: "We use Marvelous because we wanted a faster, better solution for doing realistic clothes than trying to painstakingly recreate them in Zbrush"
That does not sound like something I would have ever said... please don't tell me it's an old quote from me.
At any rate: I did not get into this business to become a freakin' tailor and I think my workflow is cluttered enough as it is without that app, hence bring on the scanners!
Replies
As far as I know the scanned data was primarily used as reference to speed up modelling. The texture resolution and quality was nothing you could simply use. In my experience hand scanners aren't that useful for textured assets. They have a very high (3d)resolution but require a static objects and the texture output (if there is any) lacks compared to high end cameras. They sit a bit between a single-camera photogrammetry workflow and a photogrammetry scan box (many cameras)
Good photogrammetry scans on the other hand are already so good you can use a lot more of the scan result as direct input for the final asset. Though most will post-process them anyway to archive a certain effect or match the assets with the pipeline requirements
But you need a very expansive scan-box with a lot of cameras + software (not that expensive anymore)
https://ten24.info/sample-scan/
I love this stuff but it seems we're moving into a direction where assets become more flexible to use instead of just single very high resolution unique textures (at least for games where you always have hardware limits).
@thomasp
wait.. I get a deja vu! Haven't I red this some time ago - just slightly different: "We use Marvelous because we wanted a faster, better solution for doing realistic clothes than trying to painstakingly recreate them in Zbrush"