[theora] No quality advantage with two-pass?

[Gregory Maxwell]

On Sat, Aug 29, 2009 at 4:38 PM, David Kuehling<dvdkhlng at gmx.de> wrote:
>>>>>> "Remco" == Remco  <remco47 at gmail.com> writes:
>> * Single-pass encoding with a constant file size will reduce quality
>> for harder parts.
>
> The problem is: there is actually no such thing as encoding for
> 'constant file size'.  there's only encoding for 'constant bit rate',

You can make the buf-delay effectively infinite as it is by default
for two-pass.

The problem is that you won't have enough 'bits' left for late complex
parts because you spent them giving early easy parts too many bits.
Two pass encoding addresses this.

Constant file-size/bit-rate must vary the quality to achieve the
desired size and buffering requirements as a result it must decide
*where* to vary the quality. One-pass must do this with no knowledge
of the future footage.

Constant 'quality' rendering has none of these problems. It encodes
the whole thing with one quality setting and the resulting size and
buffering requirements are what they are. No pesky decisions to get
wrong.