TS and Comskip

[Reddwarf]

Its not on the top of my priority list though, with so much other stuff still to be done in NPVR, and with skipping being "good enough" for most manual skipping.

I think the safest way to solve this problem would be to include the timecode in the transportstream while it is captured (http://en.wikipedia.org/wiki/MPEG_transp...m#Timecode). Is the pc's internal timecode precise enough to use?

[johnsonx42]

I think the safest way to solve this problem would be to include the timecode in the transportstream while it is captured

Perhaps this M2TS format is the answer? It looks like this is the HD format of most current digital media, so if NPVR is to be the future then perhaps M2TS (BDAV container format) is the way to go?

[sub]

I think the safest way to solve this problem would be to include the timecode in the transportstream while it is captured (http://en.wikipedia.org/wiki/MPEG_transp...m#Timecode). Is the pc's internal timecode precise enough to use?

This is an overly simplified view of a very complex problem, related to various clocks with transport streams created by digital broadcasters.

[johnsonx42]

why do we always forget you know a hell of a lot more about digital media than we do?

[Reddwarf]

But is it possible to include the internal time on each packet? The time doesn't need to be exactly correct as long as is consistent.

[GSpanFan]

In my experience dealing with TS files over the last year, I have had nothing but trouble trying to get the transport stream to play nice with anything. In my case, I wanted to get my recorded TS files to stream to my PS3 which often indicated the files were corrupt depending on the channel of recording. I was only able to do this using the ffmpeg post-processing line used above, although it only worked with an old version of ffmpeg and not newer versions. Go figure. My attempts to tinker with comskip have been fruitless, since the TS file structure appear too irregular to process.

I think what we really need is a utility that de-crapifies the captured TS stream and converts it to some standard format while maintaining audio sync and so forth. I've done a bunch of searching and the most common suggestions (TS Doctor, TSRemuxer, and others) don't seem to do it well or reliable enough. Perhaps my understanding is incorrect, but I think until this is done, the garbage in, garbage out problem will continue with any kind of post-processing.

[mark_lt]

The only clumsy solution I have found to audio out of sync is using ffmpeg and adding in -async with a very large figure, such as 400000. What this does is compensate for a messy DVB-T signal by speeding up or slowing down the audio. So, you can getting annoying chipmunk voices for small segments of the recording, but such is the compromise if you want to transcode.

My timing for comskip seems to be spot on using the TS file, and I found it is more about settings for comskip than the timecoding of the recordings, so adjusting async would not be of any help.

As a meandering aside, I found that when I changed my tuner, I had the opposite, a vastly improved quality both with audio sync and comskip, purely because the tuner handled the poor reception in my area better.

[dennit]

I tried many times to switch from DVR-MS to TS in GBPVR. I always found I had to return to DVR-MS because comskip would not skip accurately for me. I still run GBPVR for that reason. Does anyone believe that NPVR and comskip works better than TS and comskip did on GBPVR?

Thanks

[sub]

Does anyone believe that NPVR and comskip works better than TS and comskip did on GBPVR?

No - its largely the same.

[dennit]

No - its largely the same.

Is it something that is inherent in the difference between TS and DVR-MS formats that means it can't be fixed/improved, or is it just something that's hard or low priority, but is possible? Another way to ask this is: is everything necessary to create the same DVR-MS file I now record stored in a TS recording (so I could build it later from the TS file), or is something missing from it (so that the DVR-MS file would have had to have been recorded live)?