Open bordernone opened 2 years ago
@AMChierici what if the start of the video has some sort of gesture. For instance, if the question is "What do you do in your free time?" and the user records a video in the following manner: shrugs, pauses for a moment, and says "I play games, watch random youtube videos, or do nothing"
I don't expect so many such cases, but we would keep anyway 1 second before the start of a verbal sound because we need it for crossing the fade-in/out transitions.
workaround in #63
Leaving the thread open so that it can be addressed efficiently in next iterations.
Skip the playback to the point where the user starts speaking.
Possible approach: Extract the audio from the video, find the time where the intensity is high for the first time, and skip to that time.
To gain maximum efficiency with this approach, perform video preprocessing before storing it in the cloud.
For existing videos, this "fix" can be applied when the video is requested.