Sorry to bother you. I replaced the 27-bone model you provided with the 22-bone Smplh model and Mixamo model. After setting up the Actor script, GenericIK script, Motion Controller, and Blueman script, the upper body became immobile, while the lower body bones could move, but the skin did not align with the bones. Then, when I increased the bones back to 27, the whole skeleton moved correctly, but the skin still didn't align with the bones. Could you please explain what might be causing this?
This is What I want for my own UE VR system that I'm developing, instead of having to A mix OF animation & procedural foot placement which can look robotic. The bending of the Knees in the correct directions here is very impressive because dynamic crouching can be a challenge
I'm literally speechless. No joke. I don't know what to say. This is amazing. I'll look into the repo for more info but I have a use case in mind where this would be so valuable when trained on a custom dataset. Thanks Starke and everyone involved for brining us one step closer to a thriving future.
Hi! Great work. Is there any documentation, papers on this example, or information we can read for the method of this particular implementation? Cheers.
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 :)
Great! I think your work’s direction is valuable should deserve more attention. Seems like periodical modeling or idea of using fft worked really good. I think it makes sense, since bone positions naturally repositions to origin, it’s very reasonable to model it periodically. Very good xD.
This is so awesome! 😍 Is there any chance that indie developers like me will be able/allowed to use this technique and animation data set for commercial to an affordable price some day?
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.
Amazing paper! Great progress over the years! Do you think Phase embeddings could be applied to another dataset format such as audio inputs? Also, the ball has physics but what about the character? The question is: Would this simulated motion be applicable to robotic humanoids?
This is all incredibly impressive but I still can't stop staring at the red trunk dude's junk. He's got a whole gonzo nose situation. You can see it from behind!
Hello Mr Starke, I am really fascinated with applying AI and game engines together. I am far from your skill level and knowledge, but I was wondering If you could guide me on what I need to do to get into this field. I would assume I would have to take courses in Numerical Analysis, Computer Vision, and Computer Graphics. But I feel like these might not be enough. Would you be able to recommend other courses that I should take to get even an "inch" closer to your abilities/knowledge?
Machine Learning knowledge would also be needed, and in turn you would need a decent understanding of Linear Algebra and some basic calculus. Ideally you would enroll in a complete Computer Science program with some focus on AI! But of course this is state-of-the-art stuff, post-doc level, so you'd be expected to first take on some Masters and Doctorate level research.
