From: Robert R. <je...@pl...> - 2004-12-03 00:22:27
|
On Thu, Dec 02, 2004 at 02:01:39PM -0800, Scott T. Smith wrote: > On Thu, 2004-12-02 at 13:52, Robert R. Wal wrote: > > I think I know what he means. When you unauthor DVD and reencode it with > > e.g. ffmpeg, I-frames will be in different places and so chaptermarks won't > > point to I frames. Some dvd players deal with it, some don't. If you could > > unauthor DVD with each CELL in different file, you could process them > > separatelly and chaptermarks would still point to I frames (as a side > > effect). Thanks for help Robert =o) > But dvdauthor automatically makes chapters point to I frames (it takes > the user input as more of a "suggestion"). As it is supposed to. Chapter must start on GOP (not only I Frame). > > Are people running into real problems with it? Or are people just > unhappy that the chapter can drift by a couple of frames? Shifting chapter point by few frames one way or another is no real problem. Say i'm purist. I think that real proper way would be to implement I Frame insertion (GOP start) in encoder. From what I know Nic (of QuEnc -- win32 encoder based on libavcodec) is willing to implement this (there is no time-frame and i have no idea if it will be 'ported' to ffmpeg). This spliting by CellID is just other (and from my point of view) easier way of doing same thing. > One problem with dividing files at cell boundaries is that each cell is > not necessarily a standalone mpeg file. I think the audio frames can be hmm didn't think of this... I've based my idea on info that VOBU is complete stream (now i think that it was in video only context) (to clarify -- complete means with all headers and such). VOBUs are to small to deal with so... > split among two sectors that cross cell boundaries. Also, I think GOP's > at the start of a cell don't have to be closed, so ripping the video > apart there might be problematic. I think that they might not be marked as such but still are closed. Outthere in win32 world there is DVD ReBuilder which re-encodes each cell individually (this was also a bit misleading for me). But only video stream. My scenario now would be to demux chosen VTSes by cell-id. Calc bitrate [1] reencode video stream from each cell, join m2v, rip audio tracks from VOBs mux audio and video and re-author using dvdauthor. If dvdunauthor would be able to help in recreating menus it would end at this step. If not there is still BatchIfoUpdate (GPLed) that can be ported to unix (this is also on my to-do list but my time and abilities are somewhat limited). > Remember, cells aren't just used for chapters. Sometimes they are just > arbitrarily inserted! I haven't figured out why authorers do that yet. Yes, i know. I'm fiddling with dvds for over a year now and learned a lot while doing it. Yet there's still lots to learn about it =o)) Scott, have you been thinking about some 'modularisation' of dvdauthor? I mean creating something oposite to libdvdread? BTW did i say Thank You for this great app? =o))) [1] a bit off-topic to this thread: i'm trying to figure out some simple bitrate calculator but i don't know how to estimate mux overhead and subtitle stream bitrate (is 10kbit/s enough for subtitle?). Can anyone help especially with this muxing overhead. Best regards, Robert -- Just a friendly Jedi Knight | I find your lack of faith Robert Ramiega je...@pl... | disturbing |