Today we've launched our high moments detection algorithm for podcast. The challenge when generating automatic teaser from audio is trying to find great moments from that audio stream that would have been picked by a human.
It's the v1 of the algorithm, but so far the results are very good and satisfying.
Vivid should officially launch in about 6-8 weeks now.
Also part of today's update is the ability to upload custom background for the video teasers.
Any feedback always appreciated.