ShadowKnight
24th September 2004, 22:42
@jdobbs
I wanted to let you know that I was having an issue with CCE failing to encode the last episode on sex and the city season 3 disk 2. I recived the error that comes up with the HBO logo screen, although the encode didn't mess up until right after the static screen. I was running OPV directly through DVD-RB (no pre-processing, as for the most part I have given it up for stability purposes :)). I solved the issue by running RB-Obt's OPV prediction with the option to "double-pass". I was hoping that you could get the method used by Robot1 and impliment those options into your OPV prediction (double pass could be an expert option). I've found it to be more accurate as so far, everytime I have used Rb-opt it has come out to the perfect size. I hope Robot1 would be willing to share that method with you, as it seems to be more accurate at prediction.
"Double pass -- enabling double pass causes RB-Opt to run prediction in two phases. In the first it selects the sample as normal and determines the "correct" Q. In the second pass it will select a different sample (of the same size and with the same number of GOPs per range) and run the prediction again, starting with the Q value it determined in the first pass. That is to say it'll run the predicted Q over a different set of frames in the segment to verify the predicted Q was correct. If the second pass finds a Q that is within 2 points of the original, it'll use the average of the the first and second Q. If the Q from the second pass is off by more than 2, RB-Opt will suggest that this particular segment isn't suited to OPV prediction."
"2) DVD-RB's prediction is sometimes off by a couple of Qs. This is inevitable -- OPV prediction is never, and can never be, perfect -- but we feel RB-Opt's method to be a little more accurate.
3) RB-Opt's version is more configurable. You can set your desired sample size, safety margin, "GOPs per range" (see below), and can even choose a two-pass mode for better accuracy."
Thanks for listening ;) your doing great work on this program, its getting better all the time. That goes for both Jdobbs and Robot1. Thanks :)
I wanted to let you know that I was having an issue with CCE failing to encode the last episode on sex and the city season 3 disk 2. I recived the error that comes up with the HBO logo screen, although the encode didn't mess up until right after the static screen. I was running OPV directly through DVD-RB (no pre-processing, as for the most part I have given it up for stability purposes :)). I solved the issue by running RB-Obt's OPV prediction with the option to "double-pass". I was hoping that you could get the method used by Robot1 and impliment those options into your OPV prediction (double pass could be an expert option). I've found it to be more accurate as so far, everytime I have used Rb-opt it has come out to the perfect size. I hope Robot1 would be willing to share that method with you, as it seems to be more accurate at prediction.
"Double pass -- enabling double pass causes RB-Opt to run prediction in two phases. In the first it selects the sample as normal and determines the "correct" Q. In the second pass it will select a different sample (of the same size and with the same number of GOPs per range) and run the prediction again, starting with the Q value it determined in the first pass. That is to say it'll run the predicted Q over a different set of frames in the segment to verify the predicted Q was correct. If the second pass finds a Q that is within 2 points of the original, it'll use the average of the the first and second Q. If the Q from the second pass is off by more than 2, RB-Opt will suggest that this particular segment isn't suited to OPV prediction."
"2) DVD-RB's prediction is sometimes off by a couple of Qs. This is inevitable -- OPV prediction is never, and can never be, perfect -- but we feel RB-Opt's method to be a little more accurate.
3) RB-Opt's version is more configurable. You can set your desired sample size, safety margin, "GOPs per range" (see below), and can even choose a two-pass mode for better accuracy."
Thanks for listening ;) your doing great work on this program, its getting better all the time. That goes for both Jdobbs and Robot1. Thanks :)