[Vorbis] Encoding multiple tiny audio files

[Gregory Maxwell]

On Mon, Mar 30, 2009 at 9:31 AM, andy <andym00 at gmail.com> wrote:
> On Mon, Mar 30, 2009 at 2:13 PM, Gregory Maxwell <gmaxwell at gmail.com> wrote:
>> Although, since you are clearly processing the audio, you should be aware that
>> lossy compression may just be generally unsuitable for your application as a
>> lossy compressor makes assumptions about how the audio will be heard that your
>> processing may violate.
>
> Will that really be a problem you think ? I'm not so familiar with the
> Vorbis psychoacoustics but I'd say the the initial state will
> something similar to as if coming from complete digital silence (or
> whatever the ATH is dictating initially)  and therefore the results
> would good, in fact I suspect this is where the higher bitrate comes
> from because it's assuming exactly this and over-encoding (in the
> context of this one isolated sound snippet).. Well this is the
> experience I had with other psychoacoustic models before, especially
> when the sounds are very very short and highly transient..

"since you are clearly processing the audio"…"that your processing may violate"

I'm making the possibly incorrect assumption that if he is chopping
the audio into 100ms segments that he is likely doing other things to
it.

Simply passing the output through a high-pass filter at one to a couple
kHz violates the model assumptions, for example, and will usuallyresult
in audible artifacts.

This is a separate issue from his efficiency problem.