Comskip vs ShowAnalyzer?

[bits]

Originally Posted by mgpaulus

Hi guys,

Resurrecting an old thread because it's old, and I'm trying to get my system set up, and comm skipping is definitely a gotta have. So, which one are people using and happy with now? Are updates still timely and accuracy still high? I would rather watch a commercial than have part of the show removed due to falsely marking a commercial (I've had that happen using mythcommflag).

Are the recordings mpeg or h264?

IMHO, which is based on years of use and a lot of trial and error, comskip is on average more accurate than SA. However, neither is perfect. Fortunately the comskip plugin has a setting that will either automatically skip commercials or not. I have mine set to manually skip each commercial and it works quite nicely.

YMMV but for our recordings the vast majority of the time the start of the commercial segment is spot on but the end of the commercial segment can be be a few secs short of the true end of the commercial. This is ok because you do not miss any of the show.

If your recordings are mpeg then commercial detection will be faster and more accurate. If they are H264 then slower and a little less accurate. I have found through trial and error and many recordings that comskip is considerably better at finding commercials than SA on H264 files. However it is slower. You will need the donators version of comskip to do H264. Make a donation and you can get donator's version updates for life. Comskip is kept more up to date than SA as well. There really are no other alternatives.

[graywolf]

I'm using Comskip but still haven't figured out a good/consistent ini settings yet.

Some shows it is spot on, others got horribly wrong. Main reason I have auto-skip off.

My playing around with the ini file goes in spurts, when I have some time to fool with it. Although still not sure which parms to set.

Anyone out there have consistent good results for mpeg (SD) US networks that would care to share their ini file so I can compare and use as a baseline?

[bits]

Originally Posted by graywolf

I'm using Comskip but still haven't figured out a good/consistent ini settings yet.

Some shows it is spot on, others got horribly wrong. Main reason I have auto-skip off.

My playing around with the ini file goes in spurts, when I have some time to fool with it. Although still not sure which parms to set.

Anyone out there have consistent good results for mpeg (SD) US networks that would care to share their ini file so I can compare and use as a baseline?

I have messed with the ini file many times and in the end the default values work on average as good any of the modified settings I tried. If you record a wide variety of show types you will find some are spot on and others are off. I do have a few of the default settings changed and will post my ini when I get home tonight.

The broadcasters seem to be constantly looking for ways to get more commercials in and making it more difficult to accurately determine the start/stop.

What concerns me is somewhat is that I was planning on going to Hulu and such and cancel cable but there are more and more commercials now in the online content with no way to get around them.

[erik]

I found that for many tuning requests I get the first setting I have to change is
min_show_segment_length=250

[themeister]

Originally Posted by mgpaulus

Hi guys,

Resurrecting an old thread because it's old, and I'm trying to get my system set up, and comm skipping is definitely a gotta have. So, which one are people using and happy with now? Are updates still timely and accuracy still high? I would rather watch a commercial than have part of the show removed due to falsely marking a commercial (I've had that happen using mythcommflag).

I've just invested a lot of time in trying to get this to work, so here is my experience...perhaps it will save you some time:

ShowAnalyzer - Downloaded and installed the latest version. It doesn't work well at all. Blue screens, hung processes, etc. Stay away (actually run away) from it unless someone out there has an earlier version that might work. All of the SA forums seem to be focused on how to tread water and get around the bugs while the developer is off developing a new product.

Comskip - Downloaded and installed latest version. It didn't work until I realized it won't read the .ini file unless it's in the right place (the capture directory worked). The instructions say to put it in the same directory as the executable....wrong. Now it works fine (so far).

Dirmon2 - Works fine for scheduling and cleanup. SJQ evidently is great; however, it appears that Sage 7 makes it easier to deal with. I'm on 6.

Also, you want to use edl files, not txt. Since I have 3 tuners, I use 3 concurrent instances of comskip and am using the live setting for comskip.

So, if I can make it a few days without process hangs and blue screens, I'll be one happy camper. Hope this helps.

--Bill

[traker1001]

Once I went to all HD, I can't get any of them to work. SA works kinda, I actually get some detection but it seems to miss about 60% of the time.

Comskip used to work reasonably well, Then I went to HD and I can get a detection about 2% of the time. I have tried every setting I have ever read and tried altering them to best of my ability and understanding. and barely get any better. Im hoping when I get some time to email the needed info, Erik can help me out.

Nice thing about comskip is Erik is a stand up guy and supports and continues to develop his stuff.

Graywolf, You know there is a way to use different INI for different channels, don't you?

On a side note, Why is comskip less accurate on H264 files, I would think the cleaner audio and video would enhance accuracy?

On another side note, Something that I have wondered is, If there were a community developed open sourced Commercial skip program, I wonder if it would be more accurate. Unfortunately I don't know enough about programming to start one to find out.
Erik, This is meant nothing against you, I still greatly appreciate all your work and you program. Just seems multiple heads are better than one in alot of cases.

[tmiranda]

I almost always use comskip because:
1 - It works very well
2 - It's supported very well by Erik

mythcommflag is the open source program used by Myth. At one point I started looking at supporting that in CommercialDetector under Linux but since comskip works so well under Linux using wine I decided not to spend the time on it.

I've also heard that comskip is more accurate than mythcommflag. I think this is a case where one focused head is better than multiple heads.

[bits]

Originally Posted by traker1001

Once I went to all HD, I can't get any of them to work. SA works kinda, I actually get some detection but it seems to miss about 60% of the time.

Comskip used to work reasonably well, Then I went to HD and I can get a detection about 2% of the time. I have tried every setting I have ever read and tried altering them to best of my ability and understanding. and barely get any better. Im hoping when I get some time to email the needed info, Erik can help me out.

Nice thing about comskip is Erik is a stand up guy and supports and continues to develop his stuff.

Graywolf, You know there is a way to use different INI for different channels, don't you?


On a side note, Why is comskip less accurate on H264 files, I would think the cleaner audio and video would enhance accuracy?

On another side note, Something that I have wondered is, If there were a community developed open sourced Commercial skip program, I wonder if it would be more accurate. Unfortunately I don't know enough about programming to start one to find out.
Erik, This is meant nothing against you, I still greatly appreciate all your work and you program. Just seems multiple heads are better than one in alot of cases.

Are you using the latest donators version of comskip? For comskip to work properly on H264 you must use the donators version. Using the latest comskip donators version I find that it is accruate 80%+ of the time just using the default settings in the ini file.

[traker1001]

Originally Posted by bits

Are you using the latest donators version of comskip? For comskip to work properly on H264 you must use the donators version. Using the latest comskip donators version I find that it is accruate 80%+ of the time just using the default settings in the ini file.

LOL, I specifically stated that it did work about 2% of the time in hopes that no one would would tell me I needed the donators version.
But yes, I did use the latest version, However, I went back about 5 versions because I get a bit better detection. Still no where near 80% though.

[bits]

Originally Posted by traker1001

LOL, I specifically stated that it did work about 2% of the time in hopes that no one would would tell me I needed the donators version.
But yes, I did use the latest version, However, I went back about 5 versions because I get a bit better detection. Still no where near 80% though.

You still have not actually said that you are using the donators version. You keep saying that you are using the latest version which could be either the non-donator or the donator version. It is an important point. Assuming you are using the donators version; then what do you think is so different about your files than mine that causes such poor detection? I am getting at least 80% accurate detection on all Colosuss recorded H264 HD shows. I record shows from a lot of different networks but I do not record Showtime or HBO because I do not have them as part of my cable package. 2% is pretty much not at all and that does not make any sense. Are you re-encoding your recordings before comskip'ing them?

When I say that I get 80% accuracy I mean that 100% of the commercials are detected in all shows(mostly HD but a few SD) but that about 20% of the commercial segments end sometime before the commercial segment actually ends. Usually about 2-4 secs premature.

I am assuming that you realize that comskip is significantly(it takes 3-5 times longer than for mpeg) slower on H264 than mpeg and that you should wait until comskip has completed the file before viewing it. SA is faster than comskip but is less accurate.

BTW did you try the settings for comskip that I posted in one of your other posts regarding comskip settings?

[traker1001]

I more or less did, as the non-donaters version would work 0percent of the time on HD.

Originally Posted by traker1001

Comskip used to work reasonably well, Then I went to HD and I can get a detection about 2% of the time.

LOL, However, Just to clarify I am using the donors version.

I have no clue what is so different about my files. My files are TS files, Encoded 3.8gb hour. My input is 2 colossus cards via component. My audio is also component at the moment. I dont use livetv encoding. I am Using DirecTV. MY cpu is currently a q6600 running at 2.8ghz with 8gb of ram. Time it takes really isn't an issue, As it runs on a headless server.

I did try the settings and they improved the accuracy a touch, But not by much at all.

I will ask erik to help as soon as I have some time to worry about it and When I find out what the difference is Ill post it here and let you know

[bits]

Originally Posted by traker1001

I more or less did, as the non-donaters version would work 0percent of the time on HD.


LOL, However, Just to clarify I am using the donors version.

I have no clue what is so different about my files. My files are TS files, Encoded 3.8gb hour. My input is 2 colossus cards via component. My audio is also component at the moment. I dont use livetv encoding. I am Using DirecTV. MY cpu is currently a q6600 running at 2.8ghz with 8gb of ram. Time it takes really isn't an issue, As it runs on a headless server.

I did try the settings and they improved the accuracy a touch, But not by much at all.

I will ask erik to help as soon as I have some time to worry about it and When I find out what the difference is Ill post it here and let you know

Maybe it is how each of us defines accuracy? When I say 80% accurate I mean that 100% of the commercial segments(this is a group of commercials) have been found and that each of those commercial segments is roughly 80% accurate. In other words the percentage of commercial segments that are perfect is pretty low or that the majority of them are not accurate but quite useable and far exceed fast forwarding.

From all the years that I have been finding commercials that is about as good as it gets.