[MEncoder-users] Different versions, very different results

[Christian Ebert]

* Grozdan on Saturday, July 25, 2009 at 11:01:36 +0200
> 2009/7/25 Matyas <mplayer.list at sustik.com>:
>> While testing encoding settings I noticed that I get quite different results
>> with different levels of mencoder.
>> 
>> 1. The current SVN fails in the second pass with:
>>  x264 [error]: bad frame number ...
>> 
>> 2. Version r28850 determines a bitrate of 734 kb/s in the first pass.
>> 
>> 3. Version r29108 determines a bitrate of 647 kb/s in the first pass.
>> 
>> All the same parms.
>> 
>> Was there a vast improvement in compression efficiency from r28850 to
>> r29108?
> 
> Compression efficiency has nothing to do with mencoder when you're
> using H.264. It depends on the libx264 library which mencoder uses. A
> few days ago, the x264 devs added a new VAQ which in many cases lowers
> the bitrate of the encoding and in some cases increases it. It is not
> enabled by default but if you used aq_mode=2  that may explain why you
> get such different bitrate results. Of course, assuming you're using
> the x264 revision which offers aq_mode=2

Ah, thanks for pointing to this new parameter. From mediainfo I
get that aq_mode=1 is the default (outputs aq=1:1.00 when I set
neither aq_mode nor aq_strength). Shouldn't that be 0 from what
you're saying? Also the manpage is misleading:

       aq_mode=<0-2>
	      Defines how adaptive quantization (AQ) distributes bits:
		 0    disabled
		 1    Avoid moving bits between frames.
		 2    Move bits between frames (by default).