I’m from the Shopify VR team and today we released our first public VR app: Thread Studio. Thread Studio lets you test t-shirt designs in VR and create print files you can turn into real-life shirts. It was an Unreal Engine 4.13 project and we had a blast making it.
Looking great, and I totally agree with the post published on Medium by your head of VR. When I was working on something similar and introduced some of the meshes from my studio shoot pack (similar to yours) the client requested that I skip that and emulate a traditional store. And this isn’t an isolated case. I think, in the beginning, a lot of people will actually feel more comfortable experiencing things that aren’t exaggeratedly out of the ordinary when it comes to VR environments.
I’m (try to) working on something similar, where you let the user manipulate (or ‘pose’) the character. Can you talk a little bit about how you guys achieved that?
I did not have a try Thread Studio yet (still waiting for my VR stuff to arrive). I wonder if you can, say, dynamically add more of the mannequins (by the user) and save & load the pose to work on later?
I didn’t implement the posing so I’ll poke the engineer who did to jump in here and offer up his lessons learned and try and provide some info. I know we did our own IK system to make it work and we also spent a lot of time trying various ways to allow people to move the mannequin. People have very different intuitions about how it should work, and of course the animators had their own ideas.
We could theoretically spawn multiple mannequins, but right now the system that wires up the card decks to the mannequins is somewhat manual in the level layout. It wouldn’t be difficult to make it dynamic. Our concern would be performance, there are a lot of polys to the shirt/mannequin and they also use a couple stationary lights in a separate lighting channel to light them up.
As for saving and loading the pose, we have something that is half-way done. We have a special mode where you could pose a mannequin in a test level, press a key and it would save out the pose to a JSON file. We use those saved files to set the pose using the body pose card decks. This system was really only done so us developers could make the poses in VR and save them, making that a user feature was partly out of scope and partly left out to keep the complexity of the app from growing too large for end-users. The hands poses were done in Maya.
Also as an aside for other developers: on steam if you right click the app -> Properties and on the beta tab switch to the “Experimental_unsupported” branch, there is a version of the app that uses a recent 4.15 promoted build using the new forward renderer. The game looks crisp and clear (including highlighting some of the rendering shortcuts we were able to get away with while using Temporal AA). I was also able to leave all settings on “cinematic” and put the AA to 4x MSAA and still get a really smooth framerate on the same machine that had various combos of low to medium rendering features on the deferred render.
As the name implies this is just an experiment and was put together with no testing or modifications to the original project.