This is amazing work. Is the goal to replace the previous techniques? Where you take photos of your face from 3 sides and a generated 3d head with blendshapes will be created. The head attached to a body is the result so that it can be used in games / experiences. Do you think that will be a possibility? Will blendshapes be possible with this technique?
Hey. Read that you were able to achieve over 200fps in some cases, this being on one a100 gpu. Would multiple gpus make a difference in the speed? It's plenty fast right now for real time generation for low quality images, but it still hits a limit at let's say 1024 by 1024.
This sounds very promising!
This is amazing work. Is the goal to replace the previous techniques? Where you take photos of your face from 3 sides and a generated 3d head with blendshapes will be created. The head attached to a body is the result so that it can be used in games / experiences. Do you think that will be a possibility? Will blendshapes be possible with this technique?
Hey. Read that you were able to achieve over 200fps in some cases, this being on one a100 gpu. Would multiple gpus make a difference in the speed? It's plenty fast right now for real time generation for low quality images, but it still hits a limit at let's say 1024 by 1024.
you should check out Deferred Diffusion for High-fidelity 3D Head Avatars it would be amazing to see the comparison
That's all cool but how do you do this? What's the site or the app?