View Full Version : Tools / Work flow to merge multiple streams?

09-07-2004, 06:26 PM
Does anyone have a tool (or work flow) to automatically merge multiple copies of a stream into a single higher quality stream?

My HR10-250 has recorded a couple copies of the same program - an original broadcast and the repeat performance later in the week. Each one has a couple errors in it that are observable during play-back. Neither copy is perfect.

I'm looking for a tool that I can just aim at the two streams that builds a good copy out of the two partials. This seems like such a common problem that I don't beleive the tool doesn't already exist - although I can't seem to find it.



09-07-2004, 06:52 PM
TyTool can attach a VOB to other VOB to create a continous playing movie DVD.

So you have bad1.ty and bad2.ty (bad1 with some bad part first and bad2 after that). Take bad2.ty and create a VOB with the good part, then bad1.ty and start the cut after the bad part (is it better if the program has commercials and you want to cut them, is harder for movie only but it can be done with FAE). Create VOB and you end with bad2.vob (that is the first part) and bad1.vob (that's the second part). Enter the create IFO files/folders, select bad2.vob as only vob, then in the menus select attach bad1.vob, and the final dvd will play both parts as it was only 1 program/movie.

09-07-2004, 09:31 PM
I understand that I could manually do this, but that's not what I'm asking.

In my experience, it's very common to have a couple little drop-outs in a DTV sourced recording and I'm looking for an automatic way to correct multiple drop-outs without doing to edits by hand. It's very possible for a tool to do this automatically - I'm just wondering if that tool exists or not.

09-07-2004, 11:15 PM
No, nothing like that exists, at least not in the TiVo-universe (AFAIK).

It seems like it would be a bit harder than you seem to think.

How would the program know which of the two recordings is "better" at any given point in time? It seems like such a subjective thing....

As an extreme example:
One of the recordings might be perfectly clean, but broadcasting dead air, when the other has just a minute amount of 'noise' on the actual show.

Or maybe there's a (glitch|dropped frame) in the middle of one of the recordings, so the program went with the other during that time, but that happened to be when that channel was duing a test of the Emergency Broadcast System...

And that's _after_ you get over the whole "no two recordings are really the same" issue, which I think would be non-trivial. Stations cut in and out of ads early, sometimes, and play different ads. That part could probably be figured out, but it would take work.

09-09-2004, 03:24 AM

It goes beyond what has been described here both other pretty much straing into the nature of the impossible.

Don't discount the nature of the mpeg-2 streams themselves in quite the way you seem to want to. Just because the streams are identical shows, even from the same channel they are not necessarily "the same".

I often record shows multiple times looking for the best quality. I.e. new eps of Stargate SG-1 are on twice in the same night, again several times the next week in a repeated pattern.

An MPEG-2 file is a series of I B and P frames in a regimented pattern. In these multiple recordings I have seen completely different patterns of frames. Sometimes only off by a few but at other times off by 40-50 frames.

By off I mean pick a clear frame that starts a scene transition. (I.e. something easily identifable...) In file one it is an I-Frame. In file 2 it is a b-frame 40 frames into that GOP. (I actually had an enterprise episode that did exactly this.) While common sense would indicate a scene transition would happen at an I-Frame, it is just not the case. GopEditor can be used to prove this to yourself with ease...

So you can't just pick 1 I-frame over another as the display image between 2 sources can and will be a different type. With the extremely variable gop size on tivos, both SA and DTivo, this problem is extreme. So honestly without re-compression you have no hope.

Now if you are recompressing it can be done after a fashion. You wind up decoding both source files, manually lining them up, and then 1 display image after another you calculate quality of each display image between the 2 sources (somehow), pick the best one, pass that into a some form of a recompression system and you get a "better quality" output.

However mpeg-2 is lossy. Look at Buffy eps from the FX channel if you want to see what low bandwidth can do to MPEG-2. Recompression in mpeg-2 of an mpeg-2 source throws away even more of the same band of high Freq data that was tossed in the first compression run. Most if not all people would find it of a visibly lower quality.

Now recompressing into mpeg-4 (divx, xvid, etc...) changes the rules a touch and since compression is done differently you don't see quite the same quality loss. A newer codec by over a decade will do that. Switch to something non-mpeg entirely and you can get even better results like the highest quality settings of the Sorenson Vision 3 codec. (This is why it is so often used for movie trailers on the net.)

My point here is simply this, since a better quality copy is what is wanted a required recompression step will not achieve that.

Like I said at the very start of this message, it runs pretty much straight into the realm of impossibility.