[MPlayer-dev-eng] Re: [patch] Documentation update following lavc internal 2-pass mode commit
Rémi Guyomarch
rguyom at pobox.com
Mon Apr 22 17:18:04 CEST 2002
On Mon, Apr 22, 2002 at 04:33:59PM +0200, Rémi Guyomarch wrote:
> Comments ?
Disregard the previous, this one fix a few wraps in the man page.
--
Rémi
-------------- next part --------------
Index: DOCS/mencoder.1
===================================================================
RCS file: /cvsroot/mplayer/main/DOCS/mencoder.1,v
retrieving revision 1.36
diff -u -r1.36 mencoder.1
--- DOCS/mencoder.1 20 Apr 2002 06:40:47 -0000 1.36
+++ DOCS/mencoder.1 22 Apr 2002 15:23:09 -0000
@@ -194,32 +194,96 @@
Available options:
help get help
- vcodec=XXX use the specified codec
- (for full list, see html doc!)
+ vcodec=XXX use the specified codec :
+ mjpeg - Motion JPEG
+ h263 - H263
+ h263p - H263 Plus
+ mpeg4 - DivX 4/5
+ msmpeg4 - DivX 3
+ rv10 - an old RealVideo codec
+ mpeg1video - MPEG1 video :)
vbitrate=XXX specify bitrate in
kbit <4-16000> or
bit <16001-24000000>
+ (warning: 1kbit = 1000 bits)
+ vratetol=XXX approximated bitrate tolerance,
+ same unit as vbitrate.
keyint=XXX interval between keyframes
- (specify in frames)
- vhq very high quality
- vme=XXX motion estimation method
- vqmin=XXX minimum quantizer
- vqmax=XXX maximum quantizer
- vqdiff=XXX quantizer difference. Limits the
- maximum quantizer difference
- between frames.
- vratetol=XXX approximated bitrate tolerance
+ (specify in frames, >300 are not
+ recommended)
+ vhq high quality mode, macro blocks
+ will be encoded multiple times and
+ the smallest will be used.
+ vme=XXX [0..5] motion estimation method :
+ 0 - no ME at all
+ 1 - ME_FULL
+ 2 - ME_LOG
+ 3 - ME_PHODS
+ 4 - ME_EPZS
+ 5 - ME_X1
+ EPZS usually gives best results,
+ but you can try X1 too. FULL is
+ very slow and the others are
+ experimental.
+ The default is EPZS.
vqcomp=XXX if the value is set to 1.0, the
quantizer will stay nearly
- constant. If it's 0.0, the
+ constant (high motion scenes will
+ look bad). If it's 0.0, the
quantizer will be changed
- to make all frames
- approximately equally sized.
+ to make all frames approximately
+ equally sized (low motion scenes
+ will look bad).
vqblur=XXX blurs the quantizer graph over
time.
0.0 : no blur
1.0 : average all past
quantizers
+ vqscale=XXX [2..31] gives each frame the same
+ quantizer (selects fixed quantizer
+ mode).
+ vrc_strategy=X [0..2] different strategies to
+ decide which frames should get
+ which quantizer.
+ v4mv 4 motion vectors per macroblock,
+ may gives you slightly better
+ quality, can only be used in HQ
+ mode and is buggy with B frames
+ currently.
+ vpass=X [1,2] select internal first pass
+ or second pass of 2pass mode.
+
+The next 3 options doesn't apply to B frames, only I & P frames :
+
+ vqmin=XX [1..31] minimum quantizer
+ vqmax=XX [1..31] maximum quantizer
+ vqdiff=XX [1..31] quantizer difference.
+ Limits the maximum quantizer
+ difference between frames.
+
+The following options only apply to B frames :
+
+ vmax_b_frames=X [0..4] maximum number of frames
+ between each I/P frame.
+ vb_strategy=X strategy to choose between
+ I/P/B frames :
+ 0 allways use max B
+ 1 avoid B frames in high
+ motion / scene change (can
+ lead to filesize miss
+ predicton)
+
+In fixed quantizer mode you can use those options :
+ vqscale, vmax_b_frames, vhq, vme, keyint
+
+You can use the following options only in pass 1 of 2-pass mode or in 1-pass mode :
+ vqblur, vqdiff
+
+You can use the following options only in pass 2 of 2-pass mode :
+ vrc_strategy, vb_strategy
+
+All other options can be used in all modes.
+
.TP
.B \-lameopts
If encoding to MP3 with libmp3lame, you can specify its parameters here, like:
Index: DOCS/encoding.html
===================================================================
RCS file: /cvsroot/mplayer/main/DOCS/encoding.html,v
retrieving revision 1.37
diff -u -r1.37 encoding.html
--- DOCS/encoding.html 20 Apr 2002 07:02:35 -0000 1.37
+++ DOCS/encoding.html 22 Apr 2002 15:23:10 -0000
@@ -168,6 +168,23 @@
-divx4opts br=1100 -oac copy -o movie.avi -pass 2</CODE>
</P>
+<P><U><B>2 or 3-pass encoding using internal libavcodec controler</B></U> :
+Optionally you can use libavcodec's internal 2 or 3-pass mode, it may gives you
+better final rate accuracy than using the external, DivX4-inspired 2-pass rate
+controler with libavcodec.<BR>
+<UL>
+<B>2-pass encoding</B> :<BR>
+<CODE>rm -f lavc_stats.txt<BR>
+mencoder -dvd 2 -ovc lavc -lacvopts vpass=1 (audio-options) -o movie.avi<BR>
+mencoder -dvd 2 -ovc lavc -lacvopts vpass=2 (audio-options) -o movie.avi</CODE><BR>
+<BR>
+<B>3-pass encoding</B> :<BR>
+<CODE>rm -f frameno.avi lavc_stats.txt<BR>
+mencoder -dvd 2 -ovc frameno (audio-options) -o frameno.avi<BR>
+mencoder -dvd 2 -ovc lavc -lacvopts vpass=1 -oac copy -o movie.avi<BR>
+mencoder -dvd 2 -ovc lavc -lacvopts vpass=2 -oac copy -o movie.avi</CODE><BR>
+</UL>
+</P>
<P><B><A NAME=2.4.3.2>2.4.3.2. Rescaling movies</B></P>
More information about the MPlayer-dev-eng
mailing list