Comskip Support Forum

I've been using the donator's comskip successfully for over a year now on HD recordings (TS format) recorded by a Hauppauge HD PVR using SageTV. I've just updated to the latest version (81_30) to get the benefits of multi-threading ( I have a quad core AMD) and now there is an approximate 2 minute discrepancy between when commercials should start and end and the comskip commercial beginning and ending marks as displayed by Sage during playback. The prior version I was using was pretty accurate, but processing was slow, hence the upgrade. I've disabled timeline scaling in Sage, but the same result. I've experimented with the prior ini file as well as the ini included with the latest release. Any suggestion?

Same problem here... maybe a bug in this version?

COuld you test build 0.81.023

I made have introduced a timeline problem with build 0.81.024

I guess I should have searched a little more. I'm having a similiar problem to yours. Except it doesn't seem to matter what type of recording it is, HD or SD. The only program that seems to match the time line is one that I had from the CW channel.

I'll have to go back and see which version of the donators comskip I have. Sounds like I might need to go back to an older one?

erik wrote:COuld you test build 0.81.023

I made have introduced a timeline problem with build 0.81.024

Tried 81.023, seems to work fine.

erik wrote:COuld you test build 0.81.023

I made have introduced a timeline problem with build 0.81.024

I'm using Windows Media Center 7 and was using 023 with fantastic results. When I went to newer releases I started seeing more commericials. I went back to '023 and things seem to have straightened out. I came to the conclusion that something had changed (with respect to detection) in the newer releases but maybe not for the better

I found what it is.
But not simple to improve

I'm confident that if anyone can do it - you can

Could you test build 0.81.031 released today and see if it is as accurate as build 0.81.023?

erik wrote:Could you test build 0.81.031 released today and see if it is as accurate as build 0.81.023?

I don't know if you're speaking to me but I'll be happy to try it and report back. However, it would be way better for the original poster to also test it out as they are using SageTV and the HD PVR and I'm using Windows Media Center w/the Colossus.

i just tried donators 81_031 and the first show i watched was perfect, even better than the 81_023 version in my opinion. what ever you did worked!

thanks for your hard work on this

Just to recap, unlike the original poster I'm using Windows Media Center 7 (Not sage) but I thought I would try out 031 for you as well. Here's what I've found so far;

(4) 60 minute back-to-back recordings of the same series resulted in processed lengths of;

0:43:06
0:43:06
0:43:06
0:43:06

So, if that one series is any indication, I'd say you're doing pretty well
I have more recordings across a wider range of channels and lengths scheduled for tomorrow so I can give you a broader result then.

Ran comskip 81_031 on two HD TS files, seemed ok on the first two, but then on the third, there were some error messages and no commercials were found. Strange, as all three recording were from the same channel. The error message reported by Sage's SJQ4 log was "Width/height changing with frame threads is not implemented. Update your FFmpeg version to the newest one from Git." The a bunch of errors, such as "decode_slice_header.....error no frame!" "Allocate_progress() overflow."

Did not know that but the solution could be simple
You have to set explicitly in comskip.ini
thread_count=1
when you are comskipping recordings with dynamically changing video sizes.

the output_default=0 command doesn't seem to work in version 81.031