Re: [mythtv-users] Commercial detection idea
Being a UK expat, I can tell you it is much worse here in Aus. You can't fall alseep to the TV here, otherwise you get woken up at every ad becuase the 'apparent' change in volume is insane. I do think this is a great idea, and because they do what they do to make the ads appear louder (this energy thing Brad mentioned), it should be programmatically relatviely easy enough...I think! Good luck with it, DaveOn Apr 8, 2005 7:30 PM, Steve Christall <[EMAIL PROTECTED]> wrote:Brad Templeton wrote:> And it might also be worth looking not just at the lulls but in sudden> changes in the mean energy.>> The rules prohibit commercials from actually being louder than the show,> but that's the peak they measure. The show uses the peak only for loud> things, like explosions, but the commercial tries to get as close to it> as it can. At least some do.>Yes, don't get me started how much they crank the volume of commercialshere in the UK.As it stands the commercial detection code in the UK (CVS as of endMar05) basically doesn't work. Maybe the channels in the US dont mix itup as much as here?Anyway back to the sound. Yes the average sound levels during the adbreaks really crank (at least in the UK). To me it appears that all ads/ promos are run through a compressor, decreasing dynamic range andincreasing the average energy much much higher that the shows. Thiscombined with statistical analysis ... eg commercials run 3-4 mins, fouror six breaks an hour, I would have thought this would work.I can't code to save myself, but I can and will have a look at theaverage levels of a statistically significant number of shows to see ifit is obvious.Stay tunedBestSteve___mythtv-users mailing listmythtv-users@mythtv.orghttp://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users-- GMAIL is 'da bomb babyYEAHI have GMail invites, if you want one, email me direct.___ mythtv-users mailing list mythtv-users@mythtv.org http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users
Re: [mythtv-users] Commercial detection idea
Brad Templeton wrote: And it might also be worth looking not just at the lulls but in sudden changes in the mean energy. The rules prohibit commercials from actually being louder than the show, but that's the peak they measure. The show uses the peak only for loud things, like explosions, but the commercial tries to get as close to it as it can. At least some do. Yes, don't get me started how much they crank the volume of commercials here in the UK. As it stands the commercial detection code in the UK (CVS as of end Mar05) basically doesn't work. Maybe the channels in the US dont mix it up as much as here? Anyway back to the sound. Yes the average sound levels during the ad breaks really crank (at least in the UK). To me it appears that all ads / promos are run through a compressor, decreasing dynamic range and increasing the average energy much much higher that the shows. This combined with statistical analysis ... eg commercials run 3-4 mins, four or six breaks an hour, I would have thought this would work. I can't code to save myself, but I can and will have a look at the average levels of a statistically significant number of shows to see if it is obvious. Stay tuned Best Steve ___ mythtv-users mailing list mythtv-users@mythtv.org http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users
Re: [mythtv-users] Commercial detection idea
On Thu, Apr 07, 2005 at 10:50:20PM -0400, Chris Pinkham wrote: > > I had a idea i thought might be worth a try. Currently as i > > understand it the commercial detection code only looks at the image's > > captured, however an easy way to detect a commercial break would be to > > "listen" to the sound. When there is no sound (or under a certain > > level) and it coincides iwth a possible break then you have an even > > This has been mentioned several times and has been on my TODO list > since I started the commercial flagging code over 2 years ago, but > I've got enough things I can do with video so I haven't gotten to > the audio yet. With the new block-based ALL method, it should be > easier to integrate, but I don't have any plans on working on it > in the short-term future. If you'd like to do any coding to > detect these 'lulls' in the audio, feel free to submit a patch on > the -dev list. And it might also be worth looking not just at the lulls but in sudden changes in the mean energy. The rules prohibit commercials from actually being louder than the show, but that's the peak they measure. The show uses the peak only for loud things, like explosions, but the commercial tries to get as close to it as it can. At least some do. ___ mythtv-users mailing list mythtv-users@mythtv.org http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users
Re: [mythtv-users] Commercial detection idea
> I had a idea i thought might be worth a try. Currently as i > understand it the commercial detection code only looks at the image's > captured, however an easy way to detect a commercial break would be to > "listen" to the sound. When there is no sound (or under a certain > level) and it coincides iwth a possible break then you have an even This has been mentioned several times and has been on my TODO list since I started the commercial flagging code over 2 years ago, but I've got enough things I can do with video so I haven't gotten to the audio yet. With the new block-based ALL method, it should be easier to integrate, but I don't have any plans on working on it in the short-term future. If you'd like to do any coding to detect these 'lulls' in the audio, feel free to submit a patch on the -dev list. -- Chris ___ mythtv-users mailing list mythtv-users@mythtv.org http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users
Re: [mythtv-users] Commercial detection idea
On Thursday 07 April 2005 18:04, Justin Hunt wrote: > I had a idea i thought might be worth a try. Currently as i > understand it the commercial detection code only looks at the image's > captured, however an easy way to detect a commercial break would be to > "listen" to the sound. When there is no sound (or under a certain > level) and it coincides iwth a possible break then you have an even > better chance that its a commercial, becuase ive never seen audio > carry from the show into commercials. > > Would this be possible, i did a bit of work on winamp visualizations > and i think it shouldnt be too difficult. Anyone care to comment? Sounds like a worthwhile idea to me. You might want to take this to the -dev list to get feedback from those who know the internals though, some of the devs don't track -users that closely. -- Jarod Wilson [EMAIL PROTECTED] Got a question? Read this first... http://catb.org/~esr/faqs/smart-questions.html MythTV, Fedora Core & ATrpms documentation: http://wilsonet.com/mythtv/ MythTV Searchable Mailing List Archive http://www.gossamer-threads.com/lists/mythtv/ pgpjxikUBsAqN.pgp Description: PGP signature ___ mythtv-users mailing list mythtv-users@mythtv.org http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users
[mythtv-users] Commercial detection idea
I had a idea i thought might be worth a try. Currently as i understand it the commercial detection code only looks at the image's captured, however an easy way to detect a commercial break would be to "listen" to the sound. When there is no sound (or under a certain level) and it coincides iwth a possible break then you have an even better chance that its a commercial, becuase ive never seen audio carry from the show into commercials. Would this be possible, i did a bit of work on winamp visualizations and i think it shouldnt be too difficult. Anyone care to comment? -- Justin Hunt [EMAIL PROTECTED] http://bd.a.la Home: 416-926-1664 x 4238 Cell: 416-779-1201 ___ mythtv-users mailing list mythtv-users@mythtv.org http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users
Re: [mythtv-users] Commercial Detection
Whoops sorry! I didn't even think before posting a spoiler. I'll watch myself in the future On Sun, 27 Mar 2005 12:03:00 -0800, Ian Forde <[EMAIL PROTECTED]> wrote: > On Sun, 2005-03-27 at 13:17 -0500, Mario L wrote: > > I was having similar problems with my new addiction Nip/Tuck. I have > > decided to just go on what commercial detection has been finding & > > just creating a cut list before I watch the show. Hell, it takes 5-10 > > minutes to put together a cutlst & the show doesnt break when suddenly > > a commercial is detected 5 seconds long when Christian and Julia start > > to make out (Megan O'hare episode), or a 3 second commercial break > > while Juila walks in on Matt's 3 some. Its probably better this way, > > it forces me to perfect all the commercial breaks before I get ready > > to burn to DVD, intead of just go on the breaks that are still a > > little bit off. > > And I'm just in the middle of watching the Megan O'Hara episode! Thanks > for the spoiler! ;) > > -I > > ___ mythtv-users mailing list mythtv-users@mythtv.org http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users
Re: [mythtv-users] Commercial Detection
On Sun, 2005-03-27 at 13:17 -0500, Mario L wrote: > I was having similar problems with my new addiction Nip/Tuck. I have > decided to just go on what commercial detection has been finding & > just creating a cut list before I watch the show. Hell, it takes 5-10 > minutes to put together a cutlst & the show doesnt break when suddenly > a commercial is detected 5 seconds long when Christian and Julia start > to make out (Megan O'hare episode), or a 3 second commercial break > while Juila walks in on Matt's 3 some. Its probably better this way, > it forces me to perfect all the commercial breaks before I get ready > to burn to DVD, intead of just go on the breaks that are still a > little bit off. And I'm just in the middle of watching the Megan O'Hara episode! Thanks for the spoiler! ;) -I ___ mythtv-users mailing list mythtv-users@mythtv.org http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users
RE: [mythtv-users] Commercial Detection
I find that it detects actual commercials quite well but sometimes does a false commercial break. For that reason I turn off automatic skip and have a button programmed on the remote to skip forward and another to skip back that way if it screws up and marks the last 10 minutes of the show I can jump back and then use the FF and RWD buttons to jump thru the bad detection. Works quite well and wife acceptance factor is 100% now that she is properly trained :) She sees the start of a commerical and loves the fact she can hit one button and skip them all. > -Original Message- > From: [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED] On Behalf Of aaron > Sent: Sunday, March 27, 2005 12:54 PM > To: Discussion about mythtv > Subject: [mythtv-users] Commercial Detection > > > Hi, > I don't mean this to be a complaint, but rather more of a > data point... > > On Jan 17, there was a fairly major revamping of the way > commerical detection works. I found that with this new code, > detection was working better than I ever imagined was possible. > > Then, on Jan 23, there was a modification to the "brightness > thresholds". With this change, the commercial detection was > still VERY good, but not as good as the Jan 17 version. > > Since then, other changes were made, including some sort of > aspect-ratio detection, I think. I've found that with the > current version, detection isn't as good as it was even prior > to Jan 17. > > It works really well on some shows, especially ones that are > letterboxed, but on others it looks almost random sometimes. > A couple of shows that used to detect really well no longer > detect well. For example, Jeopardy (you can stop laughing now > :) . The first couple of breaks detect fine, but the Final > Jeopardy round is detected as a commercial now so that, > essentially, the show ends after Double Jeopardy. It does > this consistently now, night after night. With the "old" > version of the commercial detection, all the commercials were > detected correctly. > > I don't know what to suggest for how to improve it. I'm sure > the current version is working great for some people. But for > me, it's not as good as it used to be. And it's not as good > as I know it can be. > > -- > aaron > ___ > mythtv-users mailing list > mythtv-users@mythtv.org > http://mythtv.org/cgi-> bin/mailman/listinfo/mythtv-users > ___ mythtv-users mailing list mythtv-users@mythtv.org http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users
Re: [mythtv-users] Commercial Detection
I was having similar problems with my new addiction Nip/Tuck. I have decided to just go on what commercial detection has been finding & just creating a cut list before I watch the show. Hell, it takes 5-10 minutes to put together a cutlst & the show doesnt break when suddenly a commercial is detected 5 seconds long when Christian and Julia start to make out (Megan O'hare episode), or a 3 second commercial break while Juila walks in on Matt's 3 some. Its probably better this way, it forces me to perfect all the commercial breaks before I get ready to burn to DVD, intead of just go on the breaks that are still a little bit off. On Sun, 27 Mar 2005 12:54:07 -0500, aaron <[EMAIL PROTECTED]> wrote: > Hi, > I don't mean this to be a complaint, but rather more of a data point... > > On Jan 17, there was a fairly major revamping of the way commerical > detection works. I found that with this new code, detection was > working better than I ever imagined was possible. > > Then, on Jan 23, there was a modification to the "brightness > thresholds". With this change, the commercial detection was still VERY > good, but not as good as the Jan 17 version. > > Since then, other changes were made, including some sort of > aspect-ratio detection, I think. I've found that with the current > version, detection isn't as good as it was even prior to Jan 17. > > It works really well on some shows, especially ones that are > letterboxed, but on others it looks almost random sometimes. A couple > of shows that used to detect really well no longer detect well. For > example, Jeopardy (you can stop laughing now :) . The first couple of > breaks detect fine, but the Final Jeopardy round is detected as a > commercial now so that, essentially, the show ends after Double > Jeopardy. It does this consistently now, night after night. With the > "old" version of the commercial detection, all the commercials were > detected correctly. > > I don't know what to suggest for how to improve it. I'm sure the > current version is working great for some people. But for me, it's not > as good as it used to be. And it's not as good as I know it can be. > > -- > aaron > ___ > mythtv-users mailing list > mythtv-users@mythtv.org > http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users > ___ mythtv-users mailing list mythtv-users@mythtv.org http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users
[mythtv-users] Commercial Detection
Hi, I don't mean this to be a complaint, but rather more of a data point... On Jan 17, there was a fairly major revamping of the way commerical detection works. I found that with this new code, detection was working better than I ever imagined was possible. Then, on Jan 23, there was a modification to the "brightness thresholds". With this change, the commercial detection was still VERY good, but not as good as the Jan 17 version. Since then, other changes were made, including some sort of aspect-ratio detection, I think. I've found that with the current version, detection isn't as good as it was even prior to Jan 17. It works really well on some shows, especially ones that are letterboxed, but on others it looks almost random sometimes. A couple of shows that used to detect really well no longer detect well. For example, Jeopardy (you can stop laughing now :) . The first couple of breaks detect fine, but the Final Jeopardy round is detected as a commercial now so that, essentially, the show ends after Double Jeopardy. It does this consistently now, night after night. With the "old" version of the commercial detection, all the commercials were detected correctly. I don't know what to suggest for how to improve it. I'm sure the current version is working great for some people. But for me, it's not as good as it used to be. And it's not as good as I know it can be. -- aaron ___ mythtv-users mailing list mythtv-users@mythtv.org http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users