Another Step Towards AI Generated Video
We’re making more inroads towards infinitely automatically generated video content that can match exactly the viewer’s watch. You can check out the latest Two Minute Papers here:
With some of the systems developed, we can take text and get both video AND audio, further reducing the workload needed to create new content. The new techniques can also take audio and text and then create a scene around them, or an image.
The three papers presented are here:
CoDi : Generated Anything from Anything All at Once — https://codi-gen.github.io/
Toward High Quality Contact Sound: (this paper is actually 10 years old!) https://www.cs.cornell.edu/projects/Sound/mc/
Shortest Path to Boundary for Self-Intersecting Meshes: https://arxiv.org/abs/2305.09778
While fun YouTube channels like Corridor Crew start to different generative AI techniques for creation of videos, more creators are likely to use easier tools to create compelling and entertaining videos. Then the floodgates will really open when tuneable videos (character dialogue that can adapt) will become possible.
The only issue is that sharing of videos might need to be updated to sharing the experience, as no one will be watching the exact same stream.
