Utility: Comskip

[Stuntman]

Originally Posted by erik

I did not receive any notice that they are no longer needed.

Ok.. I'll leave them in then.. what exactly are those fixes fixing?

[erik]

One is to take care of 60fps material, the other because of some rounding errors in time code calculation (if I remember well)

[jkohn]

Originally Posted by erik

I did not receive any notice that they are no longer needed.

I thought they were just for text output though, not EDL.

[erik]

Indeed.

[malore]

erik, thanks for all your work on comskip. I've been using your version for the last few months and the results have been consistently good.

[erik]

Thanks! Its a pleasure to serve all of you!

[erik]

Finally I found some time to speed up Comskip
Here are the results

Changes in 0.79.91
- MAJOR SPEED INCREASE! 20% to 45% faster (depending on your CPU) then the previous builds

[roxy99]

Wow great news Erik. Can't wait to try it!

PS- What development platform do you use if u don't mind my asking?

[erik]

VC++ 6.0
All my programming is in C.

[snipeman]

Thanks a lot - appears to be working great for me so far.

[hchucky]

Erik,

I have been noticing some strange behavior for a while on long shows. In general I think shows that are over 1 hour are affected but perhaps it is just more noticeable. In the attached example (.bad file) comskip properly detects commercials for about the first hour and then doesn't detect any after that. I have it setup to detect in real time. I reran comskip after the show was over and it appeared to detect commercials correctly for the first and second hour of the show. See the attached .log file for the results.

An interesting thing to note is that the .bad file says "151896 Frames Processed" and the good file says "269691 Frames Processed" It appears to have processed almost twice as many frames. Could there be an issue with trying to detect commercials while the show is still recording? FWIW comskip is being kicked off using DirMon.

HC

[erik]

This is the second time I hear about comskip ending too early on a show being still recorded.
The difference in frame numbers proves comskip assumed it was the end of the recording while in reality is was still ongoing.

How many tuners do you record in paralell?

[hchucky]

I have one OTA tuner and 3 250's so I could be recording up to 4 programs at once. Also FWIW I have DirMon2 configured so that up to two instance of Comskip can be running at the same time.

HC

[erik]

Could you test with
live_tv_retries=16
and
standoff=8000000
in your comskip.ini file?

[hchucky]

Done. I have some long recordings occurring today and tomorrow. I see how it performs and report back.

Thanks!
HC

[hemicuda]

I'm using 91 since shortly after it was released. I'm not having problems even w/ NASCAR shows. Now, it gets confused due to the complexity of those shows, but it does run the entire length set to process live. If something comes up I'll post though.

[erik]

Could you mail me a zipped .csv file and a corrected .txt file so I can have a look on how to improve the results?

[hemicuda]

Originally Posted by erik

Could you mail me a zipped .csv file and a corrected .txt file so I can have a look on how to improve the results?

I'll see if I can get a full 'debug' output either this weekend or next. How's that?

[erik]

[hchucky]

Originally Posted by erik

Could you test with
live_tv_retries=16
and
standoff=8000000
in your comskip.ini file?

Recorded a couple of 2 hour shows today and it appears fine. I'll keep an on it to see if the issue resurfaces.

Thanks!
HC