Just imagine a point and click adventure in a 3D space using these animations. I can see great use cases for almost anything demonstrated here (except for sitting :))
Yeah, I too am amused by the fact that Anubis is the test subject to run through these motions. To the original poster, was this being used for a specific Egyptian themed game?
Sebastian, this is some spectacular work!! These solutions are very elegant and silky smooth. Congrats and thank you for sharing the project. I'm going to have to dive into it shortly to check out implications and applications.
Amazing stuff! I've actually been waiting so see AI take care of movements like this, especially when walking up and down stairs, that I would love to see. Next step, do the same for clothing to avoid clipping :)
i decided to come back to this after a while and holy hell my jaw dropped when you said two hours of motion capture data. i'd forgotten just how absurd that little data is. anyways this is a fantastic paper and i can't wait to hopefully see tech like this in future games.
one minor detail that i noted especially near the end with the comparison between LSTM and NSM was that the model's hands occasionally clipped into itself, especially during certain animations. would it be possible to add something similar to the enviroment sensors to the model itself? already you guys did a great job with model clipping, but it's not perfect yet. also, how does this framework compensate (if it can) for different models? eg different height, limb length, size, etc. it would be awesome if you could resize these animations for different models and effectively use 2 hours of motion capture data for a infinite variety of bipedal models. anyways, i'll leave that there for other potential studies later on. i'm not an expert on this and certainly can't build anything like this at the moment, that was just stuff i was wondering as i watched.
So awesome it's sad it's for research only... and not really interested in jumping through hoops to use it in a commercial project... Will have to wait for and make do with Kinematika. I hope Unity can learn something from this project, or hire Sebastian to help or something. Would be awesome if Kinematika can reach this level of quality.
The side by side comparison @ 5:10 demoing the impact of the gating was night and day. Is there any intention to provide these systems as a middleware solution for unreal engine? There are very limited offerings in this arena currently other than one motion matching system. Good luck with your endeavors and thanks for sharing your work. Fascinating.
Many 3D characters come in a range of shapes and sizes, with different lengths for limbs. I wonder if there is a path that could train one set of motions then correctly apply it to all the different characters with no loss in fidelity. Hopefully a future paper :)
Notice how ever 10 years or so we get an amazing video like this about cool animation tech that 90% of us can't afford or ever get access to because we are to poor. And then the video just sits on you-tube for 10-15 years and we just keep scratching our heads and continue using our old tools. And what about Unreal? The most popular and most used engine.
This is just amazing.... when can we see this released on the unity asset store for commercial use? (The github licence is sadly only for research or education purposes)
Official: We do a lot of work with NSM and deep learning framework to make a in-game animation better. Real: Thank you for finance development a new open world furry-yiff game.
Seen this several times before and it got the same issue. The copyright and "only for research" means this project can not be used for commercial games. It often die after a short while after Unity updates and makes the research code outdated. It's really sad this happen over and over again. A quick search in both Unreal and Unity marketplace show that these kind of advanced locomotion systems don't exists there.
they have this on their copyright " Licensing is possible if you want to apply this research for commercial use. " so if anyone wants to use this for a videogame you could make an agreement
Is it just me or is the only stiff part about these AI animations... is the ankles? The ankles just seem a little stiff in their movement as if they only bend to half the flexibility they can really achieve. Other than that however, this technology is brilliant.
I see this happening probably in the next 10-15 years, but within the walls of indies and AAs as the current AAA market is way too "safe" to step into new boundaries and dependent from greediness and predatory rental service.
Cool. Still, the cloth brushes his bones every few seconds, completely destroying any immersion the whole effect, admittedly quite interesting, may have created. Cloth physics and it falling into other objects completely ruins such experiences for me but I'm afraid we're still quite far from implementing it properly.
@@sebastianstarke5023 When is it possible to see it compatible with UE 4? Say someone had more traditional moving mechanics in a game and wanted to change for this kind of system. Would it have to be remade from the ground up or could you fairly easy take out bits and pieces and retrain them?
![[SIGGRAPH Asia 2019] Neural State Machine for Character-Scene Interactions](/img/logo.svg){kind=logo}
