|
From: stefan <st...@sf...> - 2007-03-19 06:44:01
|
> First, if you are encoding for DVD do *NOT* use '-q 1 -b 9000'! Which is very mpeg2enc related. I now know several encoders (ffmpeg included), which produce a *lot* less bitrate-spikes and seem to avoid them completely in 2-pass-mode ... I have even seen some files where there have not been any spikes... > If you use "-q 3" or "-q 4" the problem appears to go away. Sorry,... not here... :-( > Ideally and eventually '-q' should go away completely. ALl that should > be specified is a desired average bitrate and a tolerance (how much > above/below the average is acceptable). > Huh? I don't regard this to be a good idea. There is no harm, if an encoder uses less bits than specified by the bitrate. Imagine a completly black frame: P+B frames could be close to zero-size (just the overhead to say the decoder "same than before"...). I think, mpeg2enc needs two bitrate-modes: 1) variable bitrate-mode: A local and short second pass is just used to chop off bitrate-spikes. That is, whenever the encoder realizes that there was a spike, it goes back some frames (a GOP?) and reencodes them with a lower q-scale again. But only for that spike... 2) average-bitrate-mode or "filesize-mode": Here it makes some sense to specify an upper-bound-tolerance, only. The encoder must not go beyond this. If not possible otherwise, then use 2 passes... This mode should fit a file as close as possible to a given file-size/average-bitrate... hmm,... seems like I must use ffmpeg, currently... cu Stefan |