Slow / delayed resume after skipping while paused

[sub]

Do you think you will be able to improve the general responsiveness of skipping forward and backward? It is really sluggish for me and a little painful to operate at the moment (almost 2 second delay after button press for video to skip ahead and continue playing).

Are you talking about while paused, or playing normally? I'm not seeing any noteable delay skipping while playing. I just checked with standard def MPEG2, highdef MPEG2, standard def H.264 and highdef H.264, and all of these skipped pretty quickly here. Aproximately quarter of a second after skipping I hear audio, and about another quarter of second I see video. I couldnt really expect it to be any quicker than this.

Are you able to reproduce your skipping problem on a short five minute recording? If so, then make the recording available for me to download (use something like megaupload.com) so I can try the same file here.

[mvallevand]

Still on GBPVR here, using dvr-ms, with no timeline problems, no skipping problems, perfect Comskip functioning... just no live TV on my client...

Do you make these posts simply to make yourself feel better? They add no value to NPVR discussions, and there are so many cases where GBPVR is inferior to NPVR.

Martin

[cncb]

Are you talking about while paused, or playing normally? I'm not seeing any noteable delay skipping while playing. I just checked with standard def MPEG2, highdef MPEG2, standard def H.264 and highdef H.264, and all of these skipped pretty quickly here. Aproximately quarter of a second after skipping I hear audio, and about another quarter of second I see video. I couldnt really expect it to be any quicker than this.

This happens when skipping ahead while playing normally. It is definitely longer than half a second for me. These are high-def ATSC/MPEG2 recordings that seem to skip ahead quickly when played back in other programs. I'll see if I can make a short recording available (although it will probably take forever for me to upload even a short one).

[ga_mueller]

Do you make these posts simply to make yourself feel better?.

Martin

Yes.... I feel better, thanks!

[cncb]

Are you able to reproduce your skipping problem on a short five minute recording? If so, then make the recording available for me to download (use something like megaupload.com) so I can try the same file here.

Uploading wasn't as bad as I thought. Here is a small 3.5-minute sample: http://www.megaupload.com/?d=CEIZY09C

The audio re-starts pretty quickly but it consistently takes over a second for the video to jump ahead/catch up.

[johnsonx42]

I find that skipping works as sub described most of the time, but sometimes it will get stuck and simply never resume playing. Skipping again will get it going.

What interesting is that as far as I can tell, NPVR thinks the file is playing fine. It keeps advancing the timeline even while the video is frozen; for example if I skip to 10:00 on the timeline and it gets stuck there, then wait 1 minute with the video frozen, then do a 10-second back-skip to get the video going, playback will finally resume at approximately 10:50 on the timeline - the timeline was advancing while the video was frozen.

Another illustration of this is if it freezes on a comskip jump; if I just leave it alone for long enough, NPVR will eventually get to another comskip jump point, do another skip, and resume playback at that point.

I myself do not see what cncb is describing here, a delay every time. For me skipping is very responsive, except the 1 or 2 times out of 10 it gets stuck.

It may be that each is a different variation of the same underlying problem.

edit: I see now that cncb has uploaded a sample; I'll download it and see if it behaves differently on my system than my usual recordings.

[sub]

Uploading wasn't as bad as I thought. Here is a small 3.5-minute sample: http://www.megaupload.com/?d=CEIZY09C

I downloaded this video and skipped forward and back through it several times. It worked pretty much like I described earlier. All of the skips had the audio start pretty much immediately, and the video typically started somewhere between a quarter-to-half second later. Definitely not the 2 second delay you referred to in your earlier post.

If you can post of screenshot of your decoders, I can see if I can reproduce the same problem with your decoders.

I dont have really have any control over the way the Microsoft demux starts delivering audio packets before video packets. This relates to the audio renderer being the clock source for playback. It should still be responsive though, ie you can keep skipping without waiting for video to appear.

[cncb]

If you can post of screenshot of your decoders, I can see if I can reproduce the same problem with your decoders.

I attached a screenshot of the decoder settings. Nothing special - I use LAVAudio and with bitstream output over HDMI. I also get the same skipping behavior with another PC using AC3Filter and analog audio output. Thanks for looking into this.

[sub]

I cant really see any reason why it'd be slow skipping for you. I've made a slight tweak to the next release to try to align seeks to just before start start of an iFrame. It doesnt really seem to make much difference, but it might help.

[johnsonx42]

I've played cncb's sample on my system, and I do see what cncb means now. sometimes the delay between the audio starting and the video resuming is nearly two seconds; at one point after skipping whoever was speaking was able to say the phrase "Kim has still not decided" before the video resumed, and he wasn't speaking fast - that's at least a 2 second phrase. my recordings do this too when skipping, it just seemed normal to me.

it seems clear that this is simply the video decoder waiting for an I-frame before outputting frames again- the alternative is an ugly bunch of mpeg-2 blocks as it applies new P-frames to the last I-frame from elsewhere in the video. The ATI MPEG-2 decoder used to do this, and I disliked it immensely (though I never had to guess which video decoder was in use!).

I look forward to trying the next release; I've long believed skipping to I-frames would improve matters.