Comskip Support Forum

Ok.. so I've been using the GA release of .88 for a few days.

The good: Seems to get to the end of shows when using live TV, and not crashing with Panic volume buffer errors.

The bad: Seems to be slightly less accurate for me. Maybe it's just having issues scanning with live tv, though. I am using .88 using 10 minute delayed start. .65 was using 30 minute delay.

Also, I found one program where .88 was continuously running for 2+ days. The log file just kept growing. It was at 53MB. I had to End Task on Comskip.

Torn between .88 and .65.

I noticed that with LiveTV processing, I'm getting the following errors. Is this OK? I can see it sporadically throughout the log file. The show eventually gets processed. Maybe this is contributing to the inaccuracy compared to previous version?

Retry=0 at frame=72147, time= 1204.17 seconds
[mpeg2video @ 04c9b740] Invalid frame dimensions 0x0.
[mpeg2video @ 04c9b740] Invalid frame dimensions 0x0.
[mpeg2video @ 04c9b740] Invalid frame dimensions 0x0.
[mpeg2video @ 04c9b740] Invalid frame dimensions 0x0.
[mpeg2video @ 04c9b740] Invalid frame dimensions 0x0.
[mpeg2video @ 04c9b740] Invalid frame dimensions 0x0.
[mpeg2video @ 04c9b740] Invalid frame dimensions 0x0.
[mpeg2video @ 04c9b740] Invalid frame dimensions 0x0.
[mpeg2video @ 04c9b740] Invalid frame dimensions 0x0.
[mpeg2video @ 04c9b740] Invalid frame dimensions 0x0.
[mpeg2video @ 04c9b740] Invalid frame dimensions 0x0.
[mpeg2video @ 04c9b740] Invalid frame dimensions 0x0.
[mpeg2video @ 04c9b740] Invalid frame dimensions 0x0.
[mpeg2video @ 04c9b740] Invalid frame dimensions 0x0.
[mpeg2video @ 04c9b740] Invalid frame dimensions 0x0.
[mpeg2video @ 04c9b740] Invalid frame dimensions 0x0.
[mpeg2video @ 04c9b740] Invalid frame dimensions 0x0.
[mpeg2video @ 04c9b740] Invalid frame dimensions 0x0.
[mpeg2video @ 04c9b740] Invalid frame dimensions 0x0.
[mpeg2video @ 04c9b740] Invalid frame dimensions 0x0.
[mpeg2video @ 04c9b740] Invalid frame dimensions 0x0.
[mpeg2video @ 04c9b740] Invalid frame dimensions 0x0.
[mpeg2video @ 04c9b740] Invalid frame dimensions 0x0.
[mpeg2video @ 04c9b740] Invalid frame dimensions 0x0.
[mpeg2video @ 04c9b740] Invalid frame dimensions 0x0.
[mpeg2video @ 04c9b740] Invalid frame dimensions 0x0.
[mpeg2video @ 04c9b740] Invalid frame dimensions 0x0.
Strange audio pts step of 1199.90251 instead of 0.00000 at frame 72147
Strange video pts step of -3.26943 instead of 0.01718 at frame 72147

No Logo in frames 29371 to 72570, restarting Logo search.

This indicates corrupt input to the decoder in Comskip.
There are many possible causes: Bad input signal to your decoder, bad switching in the broadcast studio, etc...
Normally this should not cause big problems as the timeline errors should automatically be corrected by comskip.

OK... Did some more playing around.

I noticed that a program I was watching with Live processing with .88, wasn't skipping commercials that well.

After the program was done recording, I saved the XML and Log. Then I tried re-processing manually by dragging to comskip .88. The XML output was the same from the first time during live_tv and afterwards manually. So I figured there must not be any difference between live processing and processing when the recording was done.

So maybe .88 was just worse at detecting commercials that .61. So I tried dragging the video to comskip .61. XML turned out the same. So I went and tried watching the program, now that it was done recording and the skips were actually much better.

So... For pre-recorded.. Seems like .88 is the same as .61. Both pretty good.

I'm wondering what happened though. While live processing is happening is the XML different, and then it gets re-written when the file is complete? That's the only thing I can figure. Either that or the DTB plug-in isn't using the XML correctly for livetv.

Any ideas? I guess I can check the XML during live processing and see if it changes afterward.

Live processing algorithm in .088 is very different from 0.61
Final results should be comparable

Does the xml get rewritten from live, when the program is complete? Just wondering why the skipping was different?

Yes,
after the recording is finished everything is rechecked with the full algorithm and a new XML is output

Ok.. Here is a 1 hour show I just played. The show played back fine, but comskip had some kind of issue. This is with .88.

[mpeg2video @ 0e7c29c0] Invalid frame dimensions 0x0.
[mpeg2video @ 0e7c29c0] Invalid frame dimensions 0x0.
Strange audio pts step of 360.52189 instead of 0.00000 at frame 21874
Strange video pts step of -3.81998 instead of 0.01718 at frame 21874
[ac3 @ 0e7c39e0] incomplete frame
overread 19

Retry=0 at frame=22127, time= 369.50 seconds
[wtv @ 04bd1a60] timeline data missing
[wtv @ 04bd1a60] timeline data missing
[wtv @ 04bd1a60] timeline data missing
[wtv @ 04bd1a60] timeline data missing
[wtv @ 04bd1a60] timeline data missing

Comskip .log doen't go any further. There is only 1 commercial block that was found.

Here is another 30 min show, where I captured the XML 25 mins into the show (livexml) and then the final XML. There are differences... Just wondering if there is some way I can tune it to make live more accurate.

Live:
<commercial end="577.977400" start="511.077233"/>
<commercial end="679.295283" start="588.988400"/>
<commercial end="1165.297467" start="1001.734067"/>
<commercial end="1504.269433" start="1333.565567"/>
After Live:
<commercial end="9.442770" start="0.767433"/>
<commercial end="680.469604" start="518.324287"/>
<commercial end="1166.188170" start="1003.742553"/>
<commercial end="1507.462437" start="1304.509687"/>
This wasn't reflected for Live, because I captured the log before this.
<commercial end="1684.022153" start="1569.841420"/>

The short answer is "no"
The live detection algorithm is very limited due to cpu load restrictions.

mogulman wrote:Ok.. Here is a 1 hour show I just played. The show played back fine, but comskip had some kind of issue. This is with .88.

[mpeg2video @ 0e7c29c0] Invalid frame dimensions 0x0.
[mpeg2video @ 0e7c29c0] Invalid frame dimensions 0x0.
Strange audio pts step of 360.52189 instead of 0.00000 at frame 21874
Strange video pts step of -3.81998 instead of 0.01718 at frame 21874
[ac3 @ 0e7c39e0] incomplete frame
overread 19

Retry=0 at frame=22127, time= 369.50 seconds
[wtv @ 04bd1a60] timeline data missing
[wtv @ 04bd1a60] timeline data missing
[wtv @ 04bd1a60] timeline data missing
[wtv @ 04bd1a60] timeline data missing
[wtv @ 04bd1a60] timeline data missing

Comskip .log doen't go any further. There is only 1 commercial block that was found.

Any idea on this? why it stopped?

mogulman wrote:Ok.. Here is a 1 hour show I just played. The show played back fine, but comskip had some kind of issue. This is with .88.

[mpeg2video @ 0e7c29c0] Invalid frame dimensions 0x0.
[mpeg2video @ 0e7c29c0] Invalid frame dimensions 0x0.
Strange audio pts step of 360.52189 instead of 0.00000 at frame 21874
Strange video pts step of -3.81998 instead of 0.01718 at frame 21874
[ac3 @ 0e7c39e0] incomplete frame
overread 19

Retry=0 at frame=22127, time= 369.50 seconds
[wtv @ 04bd1a60] timeline data missing
[wtv @ 04bd1a60] timeline data missing
[wtv @ 04bd1a60] timeline data missing
[wtv @ 04bd1a60] timeline data missing
[wtv @ 04bd1a60] timeline data missing

Comskip .log doen't go any further. There is only 1 commercial block that was found.

Any idea on this? why it stopped?

probably because comskip crashed
Does the log file also stop in this way when you run comskip after the recording is finished?
If yes, can you make the recording available to me so I can debug this problem?

erik wrote:probably because comskip crashed
Does the log file also stop in this way when you run comskip after the recording is finished?
If yes, can you make the recording available to me so I can debug this problem?

Running it manually afterwards worked fine.