I run one test with five min videos, and the hard coded defaults did a good job.
But then I tried the original hr long videos, and the results were random. It would be great to document the assumptions (if only intended for short videos) and/or rule of thumbs for selecting better constants based on the video sizes and how off these videos may be (one of my videos had a twenty min offset).
I run one test with five min videos, and the hard coded defaults did a good job.
But then I tried the original hr long videos, and the results were random. It would be great to document the assumptions (if only intended for short videos) and/or rule of thumbs for selecting better constants based on the video sizes and how off these videos may be (one of my videos had a twenty min offset).