ghost
commented
5 years ago With the new stuff, you can set the A-B loop A point, leave B unset, and issue ab-loop-dump-cache <filename>, and it will record anything in the cache starting from the A point, including anything new added to the cache. You still have to be careful with track switching and seeking, but unlike with --record-file, you can seek around within the cache region (if it's a live stream, you can seek around as much as mpv let's you, without "losing" anything).
What is the advantage of --record-file over this? One point of --record-file was that it actually "followed" the user's seek, but in practice the result was that files got corrupted (or looked like they were). Also, the user can't actually know where the recording starts and ends. The start is typically moved up to the next keyframe (after the amount the decoder buffers, which can be easily a second), the end is cut short to the last keyframe that was found. It sucked.
stifani
Hello,
this is not a bug issue as you may have understood. This is a request to keep option --record-file in future versions of mpv (when you execute this command, a message shows that "Warning: option --record-file is deprecated and might be removed in the future ( use --stream-record or the dump-cache command).." ). So let me explain why i'm asking this, why it's important to have that option and why --stream-record or --dump-cache are not good substitutes for --record-file.
Using mpv-repl you can easily bring up the console and interact with mpv by writing simple commands to dump the stream you are watching without any problem.
If you use a command like "set stream-record" it will only dump data received from the demuxer by that moment and all you have stored in cache buffer it's ignored.
If you use a command like "dump-cache timestart timeend" you can dump some segments of the cache buffer but you have to calculate hours, minutes, seconds and I can never get it to work since it will always give this message and an empty 0bytes .
[recorder] Stream 0 has too many queued packets; dropping.[recorder] Stream 4 has too many queued packets; dropping.[recorder] Stream 0 has too many queued packets; dropping.[recorder] Stream 0 has too many queued packets; dropping.[recorder] Stream 4 has too many queued packets; dropping....[recorder] Stream x has too many queued packets; dropping....The only WORKING and good solution is to seek where you want to dump your video and while you're playing, just bring up the repl-console, write "set record-file" and it will start dumping. Obviously you don't have to seek, or change stream or do nothing like that, otherwise it will corrupt the recorded file. But that's not a problem, since I know that I can have an exact copy of that stream, in that seeked position (which is inside the cache buffer) saved to a file. That's AWESOME. Imagine if there are 30 seconds of some live stream that you just saw in a live stream and you want to go back and capture, to share with other people: using this method you can easily seeking backward and write the commad to the repl-console to save those 30 seconds and when you are done, just type 'set record-file ""' to stop recording and you can always continue to watch the live stream without interruption, having the 30 seconds portion of video dumped to your disk.
So, in conclusion --record-file is a powerful and VERY EASY option to use as you may have understood and that's why I'm asking you to keep it in mpv.
tnx