Can DVD-Lab detect I-frames?

[Colmino]

Well, this may seem like something of a dumb question. I've encoded some video using CinemaCraft encoder, and imported it into DVD-Lab. I have since scanned the entire video from beginning to end, scooting to the next I-frame. And for the entire movie - which, btw, has many examples of scene changes and the like - DVD-Lab progressed 15 frames at a time. As though the only I-frames in the entire movie were at those 15 frame intervals. This seems impossible.

I am also rather confounded by the process of setting up chapter markers, even dealing with the limitation of 15-frame intervals. I pick a series of such frames, do a frame-index on the whole movie, and then double check each chapter to make sure it's a spot I have specifically chosen. And then when I build the DVD, DVD-Lab has apparently chosen new locations for half of them! The only solution is to pick spots that are some two seconds beyond what I really want, in the hope that DVD-Lab's apparent "random inaccuracy" isn't dramatic enough to override the compromise.

Anyone know if DVD-Lab just has some sort of issue with chapters and I-frames?

[bugster]

I am not that familiar with CCE, does it hav a 'detect scene change' option. If not, or if it was not selected, that will explain why I-frames are always 15 frames apart. This is the correct GOP length for DVD.

I can't really help with the 2nd question other than to say say chapter marks must always be on an I-frame. If they are out by more than 15 frames tho this doesn't explain what you are seeing.

[awlchu]

Colmino,

This is a nagging problem with DLP. Jump back over to the DLP forum and read the DLP 'confirm bugs' sticky. You will note that Oscar is trying to smash this one before going live.

I think what is happening is this:

The smallest addressable unit in a DVD is a VOBU (vob unit). By design it must include one or more complete GOPs. I think what is happening is DLP is building the DVD from the bottom up and the I frame that you specified as a chapter point is stuck in the middle of a VOBU. This of course is illegal. So, DLP moves the chapter point +/- .5s (sometimes more) to align the chapter to a new VOBU so that it aligns to the beginning of the cell (PTT).

I have confidence DLP will fix the problem before going live.

my guess, awlchu

[Colmino]

Is there any sort of workaround for this annoying bug? This basically makes chapters in DVD-Lab worthless to me, because all of my chapters demand frame-specific precision and almost none of them land on that happy 15-frame interval. After going through the trouble of specifying a couple dozen I-frames for CCE to encode, DVD-Lab still ignores them. Grr.