[Ffmpeg-cvslog] [ffmpeg]: r5416 - in /trunk: ffmpeg.c libavcodec/snow.c tests/regression.sh

Corey Hickey bugfood-ml
Mon May 29 02:38:20 CEST 2006


lorenm at natsuki.mplayerhq.hu wrote:
> Author: lorenm
> Date: Sun May 28 14:38:10 2006
> New Revision: 5416
> 
> Log:
> Snow 1pass ratecontrol

This seems to work pretty nicely. I've only tested with one 1439-second
file, but so far I have the following observations:

- On this file, snow ends up rather short of the target bitrate for the
first pass: 679 kbit/sec vs. 800. Lavc MPEG-4 and x264 get closer, at
around 790 and 750, respectively.
- Even so, the second pass hits the target bitrate very well -- much
better than when I use vqscale for the first pass instead of vbitrate
and the target quantizer I pick results in too low a bitrate.
- Adding a third pass makes no apparent difference.


Note that I started testing this before the multiple reference frames
commit, so that could change things.

-Corey




More information about the ffmpeg-cvslog mailing list