Making MP3s, Part 1
-------------------
by Jerry Kindall <kindall@manual.com>
Although MP3 is turning into a great way to expose yourself to new
music - like the new single "Icicle" from local Michigan band
Troll for Trout, or Alan Parsons' "Dr. Evil Trance Remix" of the
title track from his new album - half the fun is in rolling your
own. Happily, there are no fewer than five separate Macintosh
applications available for creating your own MP3s. With one
exception, all of them let you encode MP3 files directly from an
audio CD - and they'll do it faster than real-time with a
reasonably speedy CD-ROM drive and processor.
<http://www.mp3.com/artists/16/michigan_rocks_99.html>
<http://www.amazon.com/exec/obidos/subst/music/download/alan-parsons/
ap-main.html>
Making MP3 files of CDs you already own and playing them back on
your own equipment is perfectly legal. Making MP3 files of music
you've created and giving them away is also legal. But uploading
and downloading "bootleg" MP3s (songs encoded from commercial
albums without the artist's or record label's permission) is
illegal. Remember, it's up to you to keep your use of MP3 players
and encoders on the light side of the Force.
We donned headphones and put together a four-minute AIFF audio
file containing several different styles of music, and next week
we'll tell you how quickly our five contenders encoded MP3 files
and how these files sounded. But first, a journey into the
psychology of sound.
**Why Encoders Matter** -- When you make a 128 kilobit per second
(Kbps) MP3 file from an audio CD, the encoded file is less than 10
percent of the size of the original, which means that the encoder
essentially discards over 90 percent of the original data. It's
been known for decades that our sense of hearing is as much
between our ears as it is in them. By taking advantage of our
knowledge of how humans perceive sound (the science of
psychoacoustics), it is possible to extract the most important
parts of an audio signal and encode them with high fidelity, using
lower fidelity for less noticeable parts of the sound, or
discarding such parts altogether. This is the basic principle
behind MP3 and other lossy audio compression schemes, such as the
QDesign Music Codec built into QuickTime.
One interesting fact about the MPEG standard (of which MP3 is only
one small part) is that the specification says nothing at all
about how an MPEG encoder should work - it only defines the format
required by the decoder. This means that developers are free to
innovate their own encoding schemes - as long as the resulting
file has the right format, it can be decoded by any MP3 player.
Competition, the theory goes, will drive developers of MP3 encoder
software to develop better and better psychoacoustic simulations.
Better encoders mean better-sounding MP3 files - and the best part
is that you don't need new playback software to enjoy the
improvement, just a new version of the file.
So, counterintuitively, the software used to create an MP3 file
can have as much or more effect on its sound quality than the
software you use to listen to it. Although some MP3 playback
programs have built-in equalizers and other enhancements to allow
you to shape the sound to your liking, all software MP3 players
sound pretty much the same with those features turned off.
The good news is that the encoders we tested produced listenable
MP3s at bitrates of 128 Kbps and higher regardless of the style of
music. Bitrate is just a fancy word for how many bits are required
to encode a second of music. The more bits you use, the less audio
information you have to throw away, and thus the better the
resulting file sounds, all other things being equal. If the
bitrate of an MP3 or QuickTime file is lower than the bitrate of
your modem (generally 56 Kbps or lower), and the planets are
aligned just right, you can actually play back the file as it
downloads. Most stereo MP3s you'll find on the Internet are
encoded at 128 Kbps or higher, which means you'll need ISDN or
better to listen to them in real-time.
In naked ear tests, you'd be hard pressed to notice any
differences between the files encoded by our selection of audio
bit-crunchers. With headphones, some minor differences become
apparent, although nothing earth-shattering was revealed until we
conducted a torture test, encoding stereo files at bitrates of 64
Kbps and lower. At this point, a number of encoding inaccuracies
(commonly referred to as "artifacts") became apparent as the
encoders struggled to decide which parts of the sound were least
important and thus disposable. It was obvious which had the best
psychoacoustic models under the hood. Tune in next week to see how
the different encoders fared in our tests, including
AudioCatalyst, SoundJam MP, N2MP3, MVP, and the free MP3 Encoder.
[Jerry Kindall is the founder of Manual Labor, a technical writing
and Web design firm specializing in the Macintosh. His music
collection includes, at last count, over 900 CDs.]
<http://www.manual.com/>
Making MP3s, Part 2
-------------------
by Jerry Kindall <kindall@manual.com>
The recent popularity of MP3 goes beyond downloading music files
from the Internet. Using MP3 encoding software, you can make MP3
files from music CDs you already own. The first part of this
article discussed the ins and outs of MP3 encoding (see "Making
MP3s, Part 1" in TidBITS-504_); this week we present the results
of donning headphones and making MP3s from five popular encoding
programs.
<http://db.tidbits.com/getbits.acgi?tbart=05636>
**Xing AudioCatalyst 2.0.1** -- AudioCatalyst was the first fast
MP3 encoder for the Mac, and the first that could encode audio
directly from CD without first saving it to your hard disk. While
the initial release didn't have all the features of its Windows
predecessor, AudioCatalyst 2.0 now enjoys parity with its Windows
sibling.
<http://www.xingtech.com/mp3/audiocatalyst/>
The program feels like a Windows port, and its options are buried
in different dialog boxes. Still, it sports a number of features
its competitors don't. For one thing, it can automatically
normalize the volume level of CD tracks before encoding them.
(Many older CDs are mastered at comparatively low levels.
Normalizing boosts the signal to take advantage of the full
available dynamic range.) It also has a function to snip silence
from the beginning and the end of a track automatically.
AudioCatalyst's panoply of features defined our expectations for
other MP3 encoders. AudioCatalyst can look up track names for
audio CDs from the Internet CD Database (CDDB), so you don't have
to name the resulting files, and it enables you to specify how you
want the files to be named (e.g., track number + song title +
artist name) and will optionally create a folder for each album
and yet another enclosing folder outside that named after the
artist.
<http://www.cddb.com/>
AudioCatalyst was the first Mac MP3 encoder to create MP3s with
the full audible frequency range from 20 Hz to 20 KHz. (Older MP3
encoders cut off frequencies at 16 KHz.) At sufficiently high
bitrates, this brings MP3 closer to CD-quality realism, although
at lower bitrates, this barely audible data can cause the
representation of the rest of the audio spectrum to suffer. Like
almost all its features, AudioCatalyst lets you turn off extended-
range encoding.
AudioCatalyst pioneered variable bitrate encoding (VBR), a feature
that automatically increases the number of bits used to encode
complicated or dense passages of music, while using a lower
bitrate for simpler passages. Standard MP3 encoding, sometimes
referred to as constant bitrate or CBR, uses the same number of
bits per second throughout the file. VBR can substantially
increase the quality of some MP3s with only a modest increase in
file size. Some older MP3 players can't play VBR files, and
neither can QuickTime 4, but most current players can handle them.
AudioCatalyst is the only program in this roundup that can MP3-
encode live audio from your computer's microphone or audio line
inputs. With the other encoders, you must first record the audio
to an AIFF-format audio file using a program like the free
Coaster, then encode that file as MP3.
<http://www.in.tum.de/~rothc/coaster.html>
If you need one of the features only AudioCatalyst provides, or if
you will be converting a whole flock of files, no other program
even comes close to offering as much functionality as
AudioCatalyst. At $30, it's price-competitive with the other
full-featured encoders in this roundup, and it's by far the most
flexible. It's also one of the fastest and produces very good-
sounding files. (In our low-bitrate torture test, it came in
second.) However, the program's user interface is unnecessarily
cluttered and complicated, so if you just want to convert a few
favorite songs to MP3 without much fuss, one of the other programs
would probably be better.
**Casady & Greene SoundJam MP 1.1** -- SoundJam MP is both an MP3
player and an encoder. It can act as an audio CD controller and
play streaming MP3 broadcasts from the Internet as well. (See
"That MP3eaceful, Easy Feeling, Part 2" in TidBITS-501_ for more
on SoundJam.)
<http://www.soundjam.com/>
<http://db.tidbits.com/getbits.acgi?tbart=05603>
SoundJam is a good choice if you want to create MP3s and listen to
them in a single program. Like its playback-only competitors
Audion and Macast, it comes with a variety of "skins" for changing
the program's appearance and supports both audio and visual
effects plug-ins. It's the only player that supports Arboretum's
Realizer plug-in, which is a fancy alternative to an equalizer
that employs psychoacoustic principles to boost the audibility of
bass on small computer speakers, enhance the stereo image, and
synthesize missing high frequencies.
Despite getting high marks for value, SoundJam's current encoder
functionality isn't competitive with the other encoders. The
program has CDDB support for automatically naming your files and
can create an enclosing folder named after the album. It also
supports optional full-frequency (20 Hz to 20 KHz) encoding and
can automatically switch out of this mode when encoding at lower
bitrates. Even when encoding the full frequency range, however,
SoundJam-encoded files sound a little soft and muffled compared to
MP3s made by other programs. (Judgments of sound quality are
extremely subjective, and there is little difference between any
of the programs we looked at for bitrates of at least 128 Kbps.
SoundJam 1.1 didn't do well on our torture test, unfortunately.)
The authors of SoundJam are aware of the product's sonic
shortcomings and are working diligently to remedy them. After we
published the first part of this roundup, Jeffrey Robbin sent us a
beta version of a new version of SoundJam. He noted that some of
the features weren't finalized, and in fact they weren't even sure
what version number it would be, but he thought we'd find the
sound quality much improved. And indeed it is. This beta version
of SoundJam MP fared much better on our low-bitrate torture test,
with very few artifacts, although it accomplished this feat by
severely restricting frequency response - the resulting MP3
sounded more like AM radio than a CD. Still, we'll take a
musically coherent but muffled MP3 over an artifact-infested one
that's almost unintelligible, and the new SoundJam gave us fewer
artifacts on the low bitrate file than all but one of the other
encoders. At more typical bitrates, the muffled character we noted
in version 1.1 was much reduced. The program has also added
variable bitrate support and a feature that lets you strip out
bandwidth-robbing inaudible frequencies below 10 Hz.
SoundJam already has the distinction of being the only MP3 encoder
that takes advantage of Apple's new Velocity Engine. On a Power
Macintosh G4, assuming you can get one, it's the fastest MP3
encoder you can buy, at least until Proteron delivers a promised
upgrade to N2MP3. If you want a good deal on a multimedia player
and encoder, SoundJam is worth checking out as it stands. The
upgrade we tested will likely render it a strong competitor on the
merits of its encoder as well.
**Proteron N2MP3** -- Although we're at a loss to explain its
name, Proteron's new MP3 encoder benefits from the most intuitive
user interface of the programs in this roundup. It's so beautiful
that it makes you wonder why every MP3 encoder doesn't work the
same way. If this program doesn't win an Apple Human Interface
Design Excellence (HIDE) award, something is seriously wrong.
<http://www.n2mp3.com/>
Here's how it works. You put an audio CD into your computer's
CD-ROM drive. As it mounts, the name of the desktop CD icon
changes to the title of the CD you just put in, thanks to a quick
CDDB look-up. You open the CD icon, and inside you find icons for
the individual songs. N2MP3 tweaks this window, too, so you can
see the title and duration of each track. To convert a song to
MP3, double-click it to save it on your desktop (or another
previously designated folder), or drag the song icon from the CD
to any folder. The N2MP3 progress window pops up and a few minutes
later, your fresh MP3 file is out of the oven. N2MP3 also provides
a convenient way to encode audio tracks on Enhanced CDs (which
don't show up on the Finder desktop as audio CDs) and uncompressed
AIFF audio files.
The encoder barely has a user interface at all - just a few
dialogs that let you choose encoding settings. Although the
settings aren't as multitudinous as those in AudioCatalyst, they
are far better organized, and aside from one minor omission, all
the essentials are there. Although N2MP3 supports full-frequency
range recording, you can't turn off the feature as you can in
AudioCatalyst and SoundJam, which hinders its encoding performance
at low bitrates.
You can choose encoding settings in a dialog that pops up at the
beginning of each encode operation, or you can choose them in the
N2MP3 Settings control panel and bypass the pre-encode dialog
entirely. This "fast track" method is the closest thing to having
MP3 encoding built into the Mac OS.
There's a unique play-during-encode feature, which of necessity
limits the program to encoding at real-time speed. For fastest
encoding, turn it off. We were slightly disappointed, however, to
discover that this feature played back the _original_ audio rather
than decoding the _compressed_ audio, so you can't hear what your
encoded file will sound like. (We were hoping it would be like the
tape monitor switch on a three-head tape deck.)
Like AudioCatalyst, N2MP3 offers variable-bitrate encoding, but
provides more control. In AudioCatalyst, you can choose only one
of five quality settings subjectively labeled from Low to High.
With N2MP3, you set the minimum bitrate using the same slider you
use to set the bitrate of a fixed-bitrate file, and then use a
second slider to tell the program how good you want the file to
sound; higher quality naturally implies additional bits. The
manual reveals that when the slider is set to Better, the encoding
bitrate for each split-second frame of the encoded MP3 file is
automatically increased until there is virtually no distortion for
that frame. As you move the slider closer to the Worse end of the
scale, N2MP3 places lower and lower limits on the number of bits
that can be added to each frame.
This is a powerful feature hidden in an obscure location and
woefully under-explained, so we'll rectify that omission here. To
make the best-sounding MP3 file the program is capable of without
wasting unnecessary bits, choose the lowest possible base bitrate
(32 Kbps) and drag the VBR quality slider to Better. Each frame of
the file will then use the number of bits required for best
results, and no more. It's a bit counterintuitive that a Better
VBR file with the slider set to 32 Kbps can be significantly
larger than one encoded with a constant base bitrate of 128 Kbps,
but no other encoder offers such an easy way to get the best sound
quality with the smallest file.
When set to its Fast mode, N2MP3 is the fastest encoder in this
roundup, beating AudioCatalyst by a few seconds when compressing a
4-minute file on our 300 MHz G3 machine at constant bitrates.
Although files encoded in this mode exhibit a slight sibilance
(exaggerated high-frequencies during "sss" sounds) compared to the
original, they are acceptable. (Proteron says that their encoder
is optimized for 160 Kbps encoding, and the sibilance all but
vanished when we tried again at that rate.) N2MP3 is significantly
slower in Best Quality mode - in fact, it was slower than all but
one of the other encoders, and that other encoder is free. In our
torture test, N2MP3 was soundly trounced by AudioCatalyst. At
ordinary bitrates (128 Kbps and above), though, N2MP3 held its
own.
**QDesign MVP 1.0** -- QDesign is no stranger to digital audio
compression; their music compression technology was deemed worthy
of incorporation into QuickTime 3 and 4. MVP is, like Casady &
Greene's SoundJam MP, intended to be a combination multimedia
player and encoder. (That's not the only thing they have in
common, since the MP3 encoder in SoundJam is licensed from
QDesign.) MVP even plays back QuickTime video and has features for
finding, downloading, and buying music.
<http://www.mvpsite.com/>
MVP's encoding options are even more limited than SoundJam's. You
get to choose the (fixed) bitrate for encoding. And that's it. MVP
does have CDDB lookup for automatic naming of files and gives you
AudioCatalyst-style flexibility in name formats, but the program
inexplicably cannot encode AIFF files to MP3, which excluded it
from our time trials. With luck, QDesign will add this invaluable
feature in the future. Files it encoded also suffered from the
same slightly "soft" sound as SoundJam, for obvious reasons.
One point in MVP's favor is that it looks really nice (nicer than
most of the "skins" available for SoundJam, Macast, or Audion,
even though you can't change MVP's appearance) with an enormous
track title display. It's also extremely simple to use and costs
only $20.
**Macromedia SWA Export Xtra & Lindvall MP3 Encoder 0.12** --
Macromedia Director's Shockwave Audio (SWA) feature enables
Director files (embedded in Web pages through the company's
Shockwave plug-in) to include streaming audio. Although Macromedia
doesn't promote the fact, SWA is essentially MP3. The SWA Export
Xtra is a plug-in for the company's SoundEdit 16 audio editor,
which costs about $300. But fear not, ye cheapskates - Johan
Lindvall has written a little application called MP3 Encoder that
supports just enough SoundEdit 16 plug-in voodoo to run the SWA
Export Xtra and to remove the SWA-specific bits of the file before
saving it. It's free, and so is the plug-in. Voila, instant free
MP3 encoder.
<http://www.macromedia.com/support/soundedit/how/shock/sound_devtools.html>
<http://www.dtek.chalmers.se/~d2linjo/mp3/mp3enc.html>
No one will mistake MP3 Encoder for AudioCatalyst. Its user
interface is almost as minimal as MVP's. You can't encode directly
from audio CDs; instead, you must use MoviePlayer or the freeware
Track Thief to create AIFF audio files, which require about 10 MB
per minute of music.
<http://www.student.nada.kth.se/~d88-bli/misc/>
The SWA Xtra lacks variable bitrate support; nor can it encode the
full audible frequency range (it only goes up to 16 KHz). And it's
slow: the two slowest times in our trials were achieved with this
software in Normal and Higher Quality mode. But it does work -
very well, in fact, despite its limited frequency response. This
encoder did better on our low bitrate torture test than any of the
other programs. And did we mention it's free?
**The Final Note** -- All of the MP3 encoders in our roundup have
at least one reason to recommend them, and all produce reasonable
files at typical bitrates. MVP plays a wide variety of multimedia
files and is the least expensive of the commercial products.
SoundJam is slightly more flexible than MVP, can play Internet MP3
streams, and has the visual bells and whistles of its playback-
only competitors. It also comes with Realizer, which can improve
sound on typical computer speakers and is attractively priced
compared to a separate player and encoder.
N2MP3 produces better-sounding files, is even more configurable,
and has a elegant and simple user interface. AudioCatalyst is
extremely configurable, very fast, and produces great-sounding
files. And the SWA Xtra/MP3 Encoder combination is free and does
very nice low-bitrate encoding.
Although we had hoped a single program would pull ahead from the
pack, it wasn't meant to be. If we're forced to pick, our vote
goes to N2MP3 for most users and AudioCatalyst for audio geeks. In
fact, our dream encoder is a cross between the two: Xing's encoder
and N2MP3's user interface, with an extra checkbox or two in the
Advanced settings to satisfy our tweaker's urge. Nevertheless, the
state of MP3 encoding on the Mac has gone from lame to robust in a
remarkably short time, and that's a credit to all the developers
involved. Try all their wares to see which suits your needs best.
You'll enjoy playing with this technology.
[Jerry Kindall is the founder of Manual Labor, a technical writing
and Web design firm specializing in the Macintosh. His music
collection includes, at last count, over 900 CDs.]
<http://www.manual.com/>