Open source machine learning for video compression

[birdie]

Details and claims

Using open source software, Collabora has developed an efficient compression pipeline that enables a face video broadcasting system that achieves the same visual quality as the H.264 standard while only using one-tenth of the bandwidth.

https://www.youtube.com/watch?v=jGQeuOacxSE

[Blue_MiSfit]

Sure, I've seen a few apps like this. They're not really sending video per-se. It's more like some image data and some data about how to move that image data around to get motion. Kind of like motion compensation taken _way_ further, but very specialized

[MasterNobody]

Originally Posted by Blue_MiSfit

Sure, I've seen a few apps like this. They're not really sending video per-se. It's more like some image data and some data about how to move that image data around to get motion. Kind of like motion compensation taken _way_ further, but very specialized

Yes, pretty much it is generated talking head (and probably can't generate anything else like for example soccer video). Also this talking head looks plastic and emotionless.

Eyes only look straight, the tongue is almost invisible.


There are no wrinkles on the forehead showing expression. And something is wrong with the teeth.

[rwill]

Reminds me of this:

https://www.matrox.com/sites/default..._datasheet.pdf

"Matrox HeadCasting™ Engine
The unique Matrox HeadCasting™ Engine is designed specifically to accelerate the animation of a high-definition 1, photo-realistic 3D version of your head in real time, while offloading your CPU."