[mythtv-users] Re: Torrentocracy Patch against Mythtv-cvs

Aaron Stewart acs at hourglassone.com
Thu Mar 3 21:59:04 UTC 2005


Not necessarily so.. Depending on reception quality (I know that my
channel 3 comes in worse than most), there may be more or less latency in
blank frame detection.  Ditto to positions that are set by user when
flagging.

On Thu, 3 Mar 2005, Brad Templeton wrote:

> On Thu, Mar 03, 2005 at 08:53:01AM -0800, Chris Petersen wrote:
> > This is pretty standard -- shows are edited such that there are
> > commercial breaks in certain places.  However, it does NOT account for
> > the commercial breaks being the same length in each region (remember,
> > commercials are regional things).
>
> Over time this is solved once you have enough people watching your
> exact profile of show.   That's easy for all the cable networks,
> everybody watches exactly the same show on Discovery Channel (even the
> same east and west.)   I think it's also true for prime time network
> programming, they are all on the same tight schedule.  If not, you need
> to find other people in your town, which will eventually happen.
>
> If you are tired of waiting for this, there are ways to make this work.
>
> You start by building a signature map for the program based on detection
> of scene transitions (key frames etc.) and all the other transitions the
> commercial flagger uses.   These transitions form a fingerprint for
> sections of the show.   Ie. this section had a 20 second scene, followed
> by a 31.5 second scene, and a 8.3 second scene etc.  Such fingerprints
> would be unique.
>
> You can also, when there is closed captioning, and it's coming from the
> network feed, just not the exact frames of the exact pieces of text.
> Highly precise.
>
> Anyway, now you record user actions not just based on the NTP time,
> but relative to these points.   Ie. "The user hit fast forward 9.3
> seconds after the frame with the closed caption text 'foo'".
>
> Now you have something that works everywhere.
>
> >
> > Also, don't forget that some people record before/after shows, and myth
> > is never 100% accurate in the time you give it (I'm set to record an
>
> Actually, my myth is _always_ 100% accurate on the time.  It is the
> networks that are off.  If you run NTP and are live on the internet,
> you will always have very very good time.
>
>



More information about the mythtv-users mailing list