Re: TECH: software for recording your voice
From: | Boudewijn Rempt <bsarempt@...> |
Date: | Monday, June 28, 1999, 17:25 |
Ok, I think this calls for a little essay about sound on computers, as I
see them. This is long, and perhaps a bit complicated, but so is the
subject.
First off, sound comes in two sorts: digitally recorded sound and
programmed sound. To the digitally recorded types belong wave files (the
ubiquitous standard), realaudio files, mp3 files and a host of others.
These generally are large files. The programmed sound types basically boil
down to one: midi (.mid files). These are not recordings, but sequences of
instructions to a sound card to play a certain note using a certain
instrument, and these are generally small files. It's a lot like those old
trick pianos that record the keys the pianist presses and replays the
sequence.
Another dichitomy is between free and closed sound types: free means
that everyone can make programs that make wav files, midi files, and
some others. Programs that make these are generally cheap, or even
free. Closed means that only one company controls the specifications,
and only that company can make software to play and record the sound.
Realaudio is an example of this. These are often painfully expensive,
and if the program doesn't work, you can't get another that does work.
Mp3 falls between: the authors come down like a ton of bricks on everyone
that tries to build a recorder, but they allow everyone to build
players. Still, free recorders are available.
Recorded sound files also come in two types: the heavily compressed
files and the uncompressed files. Wav is uncompressed, and takes about
a megabyte per second for cd-quality, although if you're satisfied with
telephone quality sound, that will drop seriously. Compressed sound takes
less space for a better quality than uncompressed sound. mp3 and realaudio
are compressed. But a cd-quality mp3 or realaudio is still double the size
of a telephone-quality wav.
Then there is the dichotomy between sound types everyone can listen to,
and sound types people generally have some difficulty with. Everyone can
listen to wav files (like Irina used, and I use in my corpus and grammar).
I'm sure that if you go to
http://www.xs4all.nl/~bsarempt/andal/languages/corpus/denden.html and
click on the sound icon you will hear something (after a considerable
wait). Everyone with a reasonable sound card and Windows or a Mac can
listen to midi files - try to double click on c:\windows\media\canyon.mid,
to see if you can. If you can't, your sound card is probably deficient,
and you've been had by the computer shop. Realaudio, mp3, au, aiff and all
the others close people out, because they don't work everywhere, cost a
lot of money or both.
A last point, and it is a subtle one, is the difference between streamed
audio and the rest. As I understand it - but I've never been able to try
it out - streamed audio, like realaudio, begins to play as soon as the
bytes enter the listeners computer over the internet. That way it looks as
if the sound file is small and quickly loaded, but it isn't, really. The
other kind can only be played if all bytes, the complete file, has entered
the listener's computer, which can take a while.
So, what to do? I wouldn't touch realaudio with a ten-foot bargepole. It's
as closed as hell, and their Unix implementation is very, very buggy (so I
can't - I'm not going to risk my banking software by going online with
Windows). Besides, in the final issue, their files aren't all that small.
If sound quality is very important, I'd use mp3 - there are still free
recording programs, or programs that convert a big wav file to an mp3, and
free players for all platforms, including windows. If all you want is some
convincing speech, then the worst wav format will do. For the second
version of fellrio, you used 22khz, 8bit mono. CD quality is 44khz, 16bit
stereo - you will be able to find those figures in your sound recording
software. I use 8khz, 8bit mono for my grammar.
We record our sounds using an old mone tape-deck I once bought for use
when learning languages (you know the drill: record your pronounciation,
listen, correct yourself, repeat), and we play it on our tape-deck, and
record it with our sound card, which is a lowly Sound Blaster PCI.
If you want to have both - a bad quality, but usable for everyone, wav
file and a good mp3, make a cd-quality mp3, mail it me, and I'll encode it
for you. Our mail server can handle e-mail up to 20 mb (although it will
take some downloading).
What's important is which sound card you have, and which recording and
playing software - the standard sound recorder included by Windows is quite
painful...
Boudewijn Rempt | http://www.xs4all.nl/~bsarempt