On Thu, 04 Sep 2003, Leonard Tulipan wrote: >> IMHO the main advantage with 2-pass encoding is, that you can hit a >> target bitrate quite easily without playing with -q and -b until you >> get the best possible result. >> >> For those who don't know what it does: >> SelectRangeEvery(100,10) selects 0-9, 100-109 and so on... >> >> If you're using the smilutils to create the video and audio >> data then simply use the "-o offset' and '-f count' options: >> >> smil2yuv -o 100 -f 10 file.dv | ... >> >> With the lav* programs it is similar: >> >> lav2yuv -o 100 -f 10 file.eli | ... >> >> > So to comment on the above two answers: > > I know what the advantages of multi-pass vbr generally are. > I can fit a 60 Min movie or a 80 Min movie on one disk, both using the > max. avaible bitrate (around 2700 for video+audio) so that the fit on > one CD-R. If I wouldn't use VBR the 60min movie would e.g. only use > 500mb and the rest of the CD-R is lost.
Well, there'S three alternatives really: * CBR: constant bitrate. You can easily encode a movie to fit perfectly on a CD/DVD-R, because there's no variation in how much space e.g. a minute of the movie needs. The disadvantage is that the quality won't be too good, because scenes that need less space to look good are encoded at the same bitrate as more demanding scenes (many details, fast moving) * VBR: variable bitrate. You giv a quality factor and a max. bitrate and the encoder varies the bitrate in order to make the whole film look as good as possible. With reasonable values fpr -q and -b you will get the most out of your movie, but it's hard to get the file size optimal. * n-pass VBR: (n-1) passes are done to get the information needed to encode the movie at optimum quality and hitting the overall target bitrate very vlosely in the last pass. Normally 2-pass is enough, although you can do up to 8 (IIRC) passes with CCE. > If I understand it correctly the quality-setting (-q) - which is > different for every movie - needs to be just right, to get the perfect > results. Finding out this is the major problem (which in fact is a > kind of manual multi-pass encoding) One could say so, but you also have to keep the bitrate in mind. -q and -b should be set in a way that the resulting max bitrate of the movie is about 20-30% higher than the average bitrate. This is necessary to give the encoder room to distribute the bitrate across the movie scenes. If your avg bitrate is very close to the max bitrate you have more or less CBR. > So all those talk of SelectRangeEvery and offset/count actually leads > towards a first-pass, where the quality-setting is determined, to then > be run on the complete video? Yep, and with SelectRangeEvery you can do the first pass (or better the first (n-1) passes, because you normally have to try a few values) on only a tenth of the movie and thus save a considerable amount of time, but still having pretty close to optimal values. regards Markus ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf _______________________________________________ Mjpeg-users mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/mjpeg-users
