Re: [Usability] New Sound Preferences and Volume Control

[Andy Owen <andy-gnome-usability ultra-premium com>]

On Wed, 2008-12-03 at 06:42 -0800, Kirk Bridger wrote:
> Put another way, isn't there some way we can just intelligently guess
> which apps should be louder?  For example, the application I'm
> currently working in versus background.  Application notification vs
> application output (like music players).  The use cases for volume
> aren't all that complex, but this discussion feels like it is becoming
> complex.

This sounds like a bad direction to go in. The problem with all this
priority based sound stuff is that it takes the one thing that everyone
understands (a volume slider) and replaces it with something where the
behaviour is totally non-obvious.

If I have my music player as priority 1, and system sounds as priority
2, then what does that even mean? My guess would be that priority 1 gets
full volume, and priority 2 gets half volume, priority 3 gets 25%
volume, etc. But I'm sure that everyone else has different theories -
which means that unless it works perfectly, and the volume manager has
the same concept of priorities to me, then it will make things harder to
use, as now to get the behaviour you want, you need to work out what it
does to your priorities to set a volume level.

To make a bit more of a point of it - it is really hard (I'm thinking
"impossible", but not saying it) for a computer to know what the
perceived loudness of some audio will be. Our perception changes with
age, and if you halve the amplitude of a wave form, instead of hearing
the same thing at a lower loudness, you lose bass and treble much more
(think of those curves that show ear sensitivity as a function of
frequency, and you'll see why). For this reason, I don't think we should
be trying to be smarter than the person sitting at the computer - my
computer even tells me that its volume is 0%, while playing back sound -
we are not nearly at the stage where we can do intelligent mixing. Given
that we have no way of knowing how loud the speakers or headphones stuff
is being played back over are, this is an impossible battle.

Instead, we should make it easy for a person to adjust the mix. This
means that immediate feedback is good (including visual feedback, but
also meaning low latency audio feedback). It means that a discoverable
interface is good, and it also means that exposing a model to the user
that they can easily understand is good - and the model of "master
volume" plus "per application volume" is quite easy to understand, and
there isn't much ambiguity.

I wouldn't consider myself to be a usability expert, but I do know
enough about sound and hearing to know that a person's ear is going to
do a much better job of setting the right mix than another application.
So any design that makes it easy for me to adjust levels gets my vote.