Starting to test Comskip with h.264 support

[gplasky]

OK-I'll get it up there tonight. Unfortunately they both are movies and around 3.9 gb in size.

Gerry

[erik]

May I suggest to find a shorter recording
Space is not the problem, the ftp server can download at 8MBit/s

[phelme]

hey erik,

so I have a request from the author of SJQ (Sage Job Queue) to have you reconsider your output codes , which I believe are the following:

ERRORLEVEL = 0 means no commercials found
ERRORLEVEL = 1 means commercials found
ERRORLEVEL > 1 means something went wrong

Apparently he looks for a return of 0 as success and anything else as failure as far as categorizing the tasks upon completion. So most of my comskip run tasks are ending up in the "failed" list as opposed to "completed" under SJQ. See this thread for details.

Not a big deal, just wanted to make you aware of it. Thanks!

[erik]

Changing the return codes is not a good idea for the other users.
Please follow the suggestion made by the author of JSQ and call comskip from a small batch file where you convert the return codes to what you would like them to be.

[phelme]

Originally Posted by erik

Changing the return codes is not a good idea for the other users.
Please follow the suggestion made by the author of JSQ and call comskip from a small batch file where you convert the return codes to what you would like them to be.

will do. thanks for the quick reply.

[Deacon Crusher]

Erik

I see in the comskip ini a couple sage specific properties.

What do these do and what should the settings look like?

I ask because I seem to routinely have comskip/sage skip just a couple of seconds to soon and get back to the show a couple of seconds late and was hopping those settings might allow me to adjust for this.

EDIT

Actually, you can forget this. I must have watched a few shows in a row that were a couple of weeks old and had used a previous version of comskip. With most current beta version all mpg files have very good comskip marks. h264 files are still more hit and miss, but getting better.

END EDIT.

All shows are recorded with Sage and all are watched via a sage extender.

Thanks

[scoful]

Erik - I don't know if this would help or not but I thought I would throw it out there.

Perhaps you could improve the processing speed and reduce the overall overhead by skipping sections of the video - say 5 or 10 seconds at a time - then look at 5 or 10 seconds - then skip again and generate a rough outline of where commercials exist - then have it go back to the 'tweens' to find the actual breaks.

Maybe you already employ something like this - haven't looked closely enough. I used to use this method for reporting from a database - kind of a fuzzy logic approach to generalizing the data into sections, eliminating larger blocks that don't have relevance and scrutinizing those that do.

This approach may prove to be less accurate - and would probably require some tuning options of it's own if it is possible (as well as the ability to disable it if someone didn't like it).

Also - does the current OTC release have H.264 support? My HDPVR's are coming in today and I really HATE commercials and LOVE comskip.......

[gplasky]

Originally Posted by scoful

Also - does the current OTC release have H.264 support? My HDPVR's are coming in today and I really HATE commercials and LOVE comskip.......

OTC does NOT have h.264. Please donate to this great program and you will receive access to the latest development thread of h.264 support by Comskip.

Gerry

[erik]

I am working on a 5 times performance improvement for h.264 processing.

[Djc208]

Originally Posted by erik

I am working on a 5 times performance improvement for h.264 processing.

While I love the work you've done right now I'll just settle for a version that can get through an H.264 file without crashing.

Mine's been generating 200 MB crash files every time it glitches in Windows. Couldn't figure out why my C drive was full till I started combing each directory and found all the dump files in the userdump folder under Windows. Something like 4+ GB worth.

[babgvant]

Originally Posted by Djc208

While I love the work you've done right now I'll just settle for a version that can get through an H.264 file without crashing.

Mine's been generating 200 MB crash files every time it glitches in Windows. Couldn't figure out why my C drive was full till I started combing each directory and found all the dump files in the userdump folder under Windows. Something like 4+ GB worth.

are you using 0.80.007? it resolved a crashing issue I was having.

[Djc208]

Originally Posted by babgvant

are you using 0.80.007? it resolved a crashing issue I was having.

No, I didn't know 007 had come out, must have been recently. I'll give it a try, thanks!


UPDATE: It looks like that may have done the trick. I installed the new version and set it to run on the sample file I sent Erik and it's already made it much further than any previous attempt. Maxed out both cores on my CPU but that's what they're there for. I'll check on it when I get back home.

[vexhold]

Where can I find a solid instructional on how to implement comskip into SageTV/SageMC?

I have seen a couple, but they are confusing and although I have completed the instructions on installation of comskip and comskip monitor, nothing seems to be happening.

Am I missing something?

[JetreL]

Hopefully this should help

http://forums.sagetv.com/forums/showthread.php?t=20358

[Taddeusz]

Since development for ShowAnalyzer seems to have stalled I've switched over to comskip today and donated to the cause. H.264 ad processing is a nice thing. For the time being I have processing of .ts files disabled because it just takes too long.

I'm curious on average how long it takes to process a 1080i video? I had it running for approximately 1hr and 45min and it had gotten about 30% through an hour recording. This is running on a 2.2Ghz Athon 64 X2.

I have it setup using Comskip Monitor and is using about 50% CPU while it's running. Although neither core is maxed out as it's running.

The 5x speedup for H.264 would certainly be a welcome change considering how long it currently takes.

[erik]

A factor 2 faster is operational and will be in next release.

[SWKerr]

Mine runs on a 1.8Ghz Core Duo and seems to take about 1.5 hours per .5 show. I tested it on a 2.8Ghz Athon X2 and it wasn't really that much faster. In practice I have found I don't care. The news it the only thing I tend to watch right away and everything else seems to be done by the time I want to watch it. I haven't noticed any performance problems recording or playing back since I started using it on the server.

I would like to know if anyone has played with the config at all. I basically just dropped in the current zip files over my previous comskip setup and everything just runs. I have not used comskip in awhile since I wanted HD more than I wanted comskip but I seem to remember it being more accurate than what I have been getting lately. Has anyone tuned the ini in the standard zip file and gotten better results. It is not bad really but I seem to recall it being pretty close to perfect before I went to the HD-PVR.

[Taddeusz]

My main problem right now is that I have a huge backlog of .ts recordings so if I just enable it for all of them then I imagine I'll have to wait months for it to go through them all. If only I had a faster processor, preferably a quad core.

[JetreL]

If you have multiple computers you could make the drive that the video is stored on shared and then mount the drive on another pc and run comskip manually.

This will allow you to spread the computing power across several processors. Good Luck!

[Beefcake550]

JetreL.....Just wanted to point out that I will be starting up this task of running comskip on multiple machines soon. I have 2 Quad core PCs that will be available for operation in about a week. I have a backlog of about 400GB of ts files from my HDPVR that need processing.

I'll let everyone know how I did it when it's done.

-Brian