Dear all,
We are using vdr (1.7.21) to record from a DVB-T2 stream. File split quota is at 200MB. The broadcast stream comes with audio packets (data) roughly 2 seconds after then the video packets. They are in perfect time stamp sync (video-audio). It is just the data packets that are late in the data stream).
Now, when VDR does the file split (after the quota has been met and a video I-Frame is about to start) the audio packets that would be in sync with the latest video frames before the split are actually being contained in the next split file.
This creates me big extra workload, I want to be able to process the split files individually, but I cannot since they don't have audio-video integrity.
What would it take to make vdr splitting smarter in this case, that is to only split the files after all the significant audio packets have been received as well?
Thanks for your attention, Rares
Am 04.09.2012 15:22, schrieb Rares Pop:
Now, when VDR does the file split (after the quota has been met and a video I-Frame is about to start) the audio packets that would be in sync with the latest video frames before the split are actually being contained in the next split file.
Thats because VDR does splitting and editing on the data stream level, not on the audio/video-track level. Merging all the file pieces together always gives the original stream, with just very minor modifications.
Editing based on time stamps would be a lot more difficult, as near the splitting point, data needs to be moved to the previous/next file, depending on whether audio is ahead/behind video. To make things more complicated, there's no guarantee that together with the I-frame also a new audio frame starts at the exact time stamp. For precise editing, an audio frame may have to be split and re-compressed.
And for playback, the same applies in reverse: For pre-buffering, the A/V-data of the file end and beginning needs to be merged with proper offset again. This is all a lot more complicated than just splitting the whole data stream as one.
For further processing, its the best to merge continuous stream parts into one big file, if the program cannot handle split files on its own. If you want to avoid the copying, take a look at vdrnfofs, a virtual file system that virtually merges VDR recordings into single mpg files.
Cheers,
Udo
Thanks for the response and the vdrnfofs tip. At the moment we're doing some post processing to deal with this, by bringing back the audio from the next file.
You were saying that VDR does splitting and editing on the data stream level, not on the audio/video-track level. Am I to understand that vdr doesn't make the files at the beginning of a video I-frame?
Kind regards, Rares
On Sep 4, 2012, at 11:55 PM, Udo Richter udo_richter@gmx.de wrote:
Am 04.09.2012 15:22, schrieb Rares Pop:
Now, when VDR does the file split (after the quota has been met and a video I-Frame is about to start) the audio packets that would be in sync with the latest video frames before the split are actually being contained in the next split file.
Thats because VDR does splitting and editing on the data stream level, not on the audio/video-track level. Merging all the file pieces together always gives the original stream, with just very minor modifications.
Editing based on time stamps would be a lot more difficult, as near the splitting point, data needs to be moved to the previous/next file, depending on whether audio is ahead/behind video. To make things more complicated, there's no guarantee that together with the I-frame also a new audio frame starts at the exact time stamp. For precise editing, an audio frame may have to be split and re-compressed.
And for playback, the same applies in reverse: For pre-buffering, the A/V-data of the file end and beginning needs to be merged with proper offset again. This is all a lot more complicated than just splitting the whole data stream as one.
For further processing, its the best to merge continuous stream parts into one big file, if the program cannot handle split files on its own. If you want to avoid the copying, take a look at vdrnfofs, a virtual file system that virtually merges VDR recordings into single mpg files.
Cheers,
Udo
vdr mailing list vdr@linuxtv.org http://www.linuxtv.org/cgi-bin/mailman/listinfo/vdr
-- This message has been scanned for viruses and dangerous content by MailScanner, and is believed to be clean.
Am 05.09.2012 11:13, schrieb Rares Pop:
You were saying that VDR does splitting and editing on the data stream level, not on the audio/video-track level. Am I to understand that vdr doesn't make the files at the beginning of a video I-frame?
The video stream gets split at a TS packet that contains an I-Frame / GOP start of the VPID video stream, thats true. But all other streams get split just at that same point in the data stream, not necessarily at any special boundaries.
Cheers,
Udo
-
Rares Pop -
Udo Richter