2005-07-30, 02:05 AM
WARNING HUGE POST. PRE Written. Anyway is this something you are thinking about Jere_Jones
Okay I just started out trying to use Comskip and SA also.
SA gives me problems with the registration even though I have a test serial. Okay, so really, I can not comment on the performance.
I know some have gotten comskip to work very well.
With default settings it does average at best for me. No big deal. I know if I were to spend time tweaking it would get better.
While I was trying stuff out a little different approach came to mind for how a Commericial detection could work. This probably take alot more processor time initially but hopefully on the backend would payoff with near perfect video detection.
First set in the process would be to make a perfect comskip file. This would require the use of something like vdubmod to pinpoint the exact moment of a commericial start and stop.
You could also design a button for watching live tv. First strike would mark a commercial break second would end it. This would give you a near perfect match that you could update later with vdub.
Okay that said Now you have a perfect match for where commercials are. This is all in the comskip file.
Next you send this comskip file to the actual program comskip 2.0 or SA 2.0 (or what ever version it would be)
Now the program has your perfect cuts for commercials. Here comes the processor demanding part.
The program now with the video file it does lots of reviewing of it and out puts several configure files that would make that video file break exactly like you want it to.
I know this is alot different then now, kind of working in opposite direction. This is all for the configuration of the little program.
So now you have have these several config files (.ini) for this specific channel and also show.
Next you can feed another file through in this manner. Now this is where you can cut down on alot of processing time from the previous step. If it matches both the channel and show(same show different episode) It will then look at the short list of .ini (config) files. So it will look through these and then it will output the ones of these that will also give you exactly like you want.
Okay now you have a baseline standard for this show. When you run the comskip or SA on another episode of the same show and same channel it will use this configure file. This should lead to very accurate results in the future. Also You can further tweak it by sending more of the same episodes through.
All that said what if the second file through is a differnt show same channel but different show or maybe altogether different channel.
Scenerio 1 same channel different show. It would first give you config files for that show. Thus for future recording should be very accurate.
But also at the same time it will merge info from first shows ini's with the seconds so basically you now get a set of configure files for this channel. These would be used in the future if a new show was being recorded on this familiar channel. Sort of a semi global configure file.
Scenerio 2 if it is an all together different file well it would first establish a set of ini's for that show on that channel but also that channel would have default ini's same as the shows. This would change once you sent through another show on this channel which would then allow for a further tweaking.
Also what will happen is a MAIN global ini will be setup. This will be a combination of all channel configs. This way in the future if say you record on a new channel an average of all the channel configs will be used.
The above proceess which would probably be a bit time consuming and should probably run on a fast computer to atleast do all the primary calculations.
Anyway once all main configures are done you can have the comskip or SA possible run with an option something like comskip $file -preview. At the end it would start a preview of the show and at the end ask if the commercial detection was accurate. IF so it would further weight the .ini configuration. May want output running at 4x 2x or some other rate. so maybe play rate would be adjustable as well.
What I have written above seems like a huge task, but I feel SA is heading towards that direction, atleast from a post I read.
Anyway I feel even though this method maybe cpu demanding would ultimately allow for excellent commercial detection (although need to develope the real logic behide it all). Also like to comment that I would indeed pay for something that would allow accurate commerical detection all the time.
Hope this was easy to understand I am tired and going to bed.
Okay I just started out trying to use Comskip and SA also.
SA gives me problems with the registration even though I have a test serial. Okay, so really, I can not comment on the performance.
I know some have gotten comskip to work very well.
With default settings it does average at best for me. No big deal. I know if I were to spend time tweaking it would get better.
While I was trying stuff out a little different approach came to mind for how a Commericial detection could work. This probably take alot more processor time initially but hopefully on the backend would payoff with near perfect video detection.
First set in the process would be to make a perfect comskip file. This would require the use of something like vdubmod to pinpoint the exact moment of a commericial start and stop.
You could also design a button for watching live tv. First strike would mark a commercial break second would end it. This would give you a near perfect match that you could update later with vdub.
Okay that said Now you have a perfect match for where commercials are. This is all in the comskip file.
Next you send this comskip file to the actual program comskip 2.0 or SA 2.0 (or what ever version it would be)
Now the program has your perfect cuts for commercials. Here comes the processor demanding part.
The program now with the video file it does lots of reviewing of it and out puts several configure files that would make that video file break exactly like you want it to.
I know this is alot different then now, kind of working in opposite direction. This is all for the configuration of the little program.
So now you have have these several config files (.ini) for this specific channel and also show.
Next you can feed another file through in this manner. Now this is where you can cut down on alot of processing time from the previous step. If it matches both the channel and show(same show different episode) It will then look at the short list of .ini (config) files. So it will look through these and then it will output the ones of these that will also give you exactly like you want.
Okay now you have a baseline standard for this show. When you run the comskip or SA on another episode of the same show and same channel it will use this configure file. This should lead to very accurate results in the future. Also You can further tweak it by sending more of the same episodes through.
All that said what if the second file through is a differnt show same channel but different show or maybe altogether different channel.
Scenerio 1 same channel different show. It would first give you config files for that show. Thus for future recording should be very accurate.
But also at the same time it will merge info from first shows ini's with the seconds so basically you now get a set of configure files for this channel. These would be used in the future if a new show was being recorded on this familiar channel. Sort of a semi global configure file.
Scenerio 2 if it is an all together different file well it would first establish a set of ini's for that show on that channel but also that channel would have default ini's same as the shows. This would change once you sent through another show on this channel which would then allow for a further tweaking.
Also what will happen is a MAIN global ini will be setup. This will be a combination of all channel configs. This way in the future if say you record on a new channel an average of all the channel configs will be used.
The above proceess which would probably be a bit time consuming and should probably run on a fast computer to atleast do all the primary calculations.
Anyway once all main configures are done you can have the comskip or SA possible run with an option something like comskip $file -preview. At the end it would start a preview of the show and at the end ask if the commercial detection was accurate. IF so it would further weight the .ini configuration. May want output running at 4x 2x or some other rate. so maybe play rate would be adjustable as well.
What I have written above seems like a huge task, but I feel SA is heading towards that direction, atleast from a post I read.
Anyway I feel even though this method maybe cpu demanding would ultimately allow for excellent commercial detection (although need to develope the real logic behide it all). Also like to comment that I would indeed pay for something that would allow accurate commerical detection all the time.
Hope this was easy to understand I am tired and going to bed.