This is insane. just replaced video codecs with a neural network. We'll all be controlling digital face puppets of ourselves on video calls in the future! 👹 /cc #ThisDoesNotExist #AvatarLand #SyntheticMedia
To be fair, this method will only work on the data domain involving faces, while H.264 will work for most domains. However, at a time when much larger % of video data is from conferencing, it makes sense to tailor compression algorithms directly for this.
Very cool technology for the future of video conferencing: instead of sending (compressed) images, just generate facial movement with GANs #thatsai
This is really cool & demonstrates the capacity for innovation to help solve emerging network management issues by actually reducing network congestion rather than arbitrarily "banning" prioritization (à la net neutrality regs) h/t
Freaking cool research from uses AI to dramatically improve video conferencing with a novel approach to video compression. This uses a GAN to reconstruct high quality images of a person's face, and just 0.1% of the bandwidth (compared to H264)
Wow, this is such a good use case for synthetic media!✨Let's send vectors instead of bitmap and win over low bandwidth during videoconferencing! It even fixes the alignment of eyes with the camera and you can possibly look like anything you want👌#NVIDIA
This is a great idea, well implemented by
If this works, it’s just nuts. Nvidia has come up with a video compression algorithm that lets you do Zoom calls with a tiny fraction of the bandwidth usually required. Could be a game-changer for people with poor internet connections. (Via !)
I may need to update my graph soon
2010s: software eats hardware 2020s: hardware eats software
Super clever #AI video compression by leverages GANs to improve image quality and efficiency but the idea behind it also opens the door for #deepfakes to enter the video conferencing arena
Replacing a video codec with a neural network that reduces bandwidth consumption by orders of magnitude.
Replacing a video codec with a neural network that reduces bandwidth consumption by orders of magnitude.
New video from $NVDA Maxine - 1) Video compression via only sending facial key points over the wire 2) 3D face alignment 3) overlay avatar.. they are pushing what you can do with edge computing!