Postby TheBigTime001 » 26 Mar 2015 01:45
Indeed, calculations from keyframes would tend to be very inaccurate. Just examining a movie's keyframes in Avidemux, there is absolutely no consistency between different medias. A DVD could range between 1 to over 10 frames between keyframes, and from the Internet, there are so many poorly encoded videos that this method of calculation is doomed to fail. You could build a frame-by-frame cache, up to x frames ago, but as has been pointed out, this has the potential to be extremely RAM-intensive. He didn't even get in to UHD video's requirements, and for good reason. That could require over 1 GB of RAM for the caching. Avidemux uses a frame-by-frame index, AFAIK, but it takes up to 30 seconds to a minute or more to load the file, depending on the size and complexity. That kind of long load time is absolutely not necessary unless the user wishes it. Maybe pre-indexing of the file is potentially an option, but it would likely not be used, as I doubt that the vast majority of people would want to wait that long for files to load in VLC. Especially because settings that aren't supposed to be on by default have an odd tendency to wind up on when users play around in the advanced settings (In my first year of VLC I made many such mistakes). So it comes down to both long loading times being necessary for any practial accuracy and potential UI issues (many settings that are in the advanced settings aren't well documented and can confuse users). Sorry for the long wall of text, it's just my 5 years of VLC experience talking.
If we can simulate real life so accurately, then isn't our reality maybe just one great big simulation as well?
Please don't use PMs for support questions.