Slashdot Mirror
Non-MP3 Codecs?
Vanth Dreadstar asks: "While
MP3 is okay, I have begun researching other codecs that would be
suitable for my home music use. Lossy codecs such as Ogg
Vorbis, AAC,
and MPC all seem to have promise, not to mention the lossless codecs
such as Shorten
(otherwise known as .SHN),
LPAC, and FLAC.
I would like to know what non-MP3 codecs people are using out there,
and why."
Another consideration is the straightforwardness of the API for the library you intend to use. Vorbis has a somewhat reasonble API with a liberal addition of quirks. Also you can easily add metadata to Vorbis files. Ever tried adding metadata to an MP3 file? ID3v1.1 is trivial but ID3v2 has a 95,000 line reference implementation. Uh? UH?
Any application has to support PCM audio also, since most music collections are primarily on CD.
I recently bought a Rio Volt MP3/WMA CD player, and compared WMA8 with VBR LAME, and LAME won hands down. Both encoders are set to come out around 128kbit, and while both of course have artifacts, the artifacts in WMA are MUCH more noticeable. I guess I'm just alot more sensitive to the type of artifacts WMA produces...
Sticking with MP3s is a no brainer unless you have to use open software for moral reasons, since Apple has enhanced MP3 encoding/decoding for AltiVec, and this is an area where those gigaFlops do wonders at quick, high-quality encodes and freeing up more CPU for your work (or the visualizer :) during playback.
"Reality is just a convenient measure of complexity" -Alvy Ray Smith
is a ftp database and crawler similar to audiogalaxy only for ogg. it would catch on in no time.
Yeah tell me about it...I produce my own tracks (mostly house music), I have a hard time sending out anything but .mp3 files. What's even worse is when people ask for stuff in RealAudio or WMA for streaming purposes. I lose so much quality (especially hi-hat loops and some portions of the basslines) that I have to re-do some of my tracks so you can actually hear certain portions.
Bear in mind that the ~4x compression rate listed for lossless compression schemes is heavily reliant on the input. Don't be surprised if you get 1.5-2.5 compression a lot of the time, and remember that there's a good chance you'll get 1:1 (or worse) compression results with a 'random' enough song file.
Obliteracy: Words with explosions
JPEG users have available to them some command line utilities that permit simple alteration of images without loss of quality, for example, rotation and flipping. Are there any similar utilities available for any of the major audio compression formats?
The reason I ask is that I have ripped a number of CDs and the volume levels vary noticibly. I like to listen to MP3s as I work, with the volume turned down far enough that I can hear the music, but any one that I'm on the phone with won't. Unfortuately, there doesn't seem to be a single setting for everything that I've ripped. While I could go back and re-rip, I'd much rather have a toolbox of useful batch utilities. Ideally, it would allow me to write, say, a Perl script that generates a histogram, checks the average and peak volume, and then tweaks a single number in the file header to force it in line with the rest of my collection.
Is this sort of thing possible?
Nothing for 6-digit uids?
In most cases, a 60kbps OGG file sounds as good as an 128k mp3. An 80k OGG is as good as 160k mp3 and half the size.
Actually, Ogg only shaves off 30-40% (still respectable, just not revolutionary)
If you have a portable player, you would appreciate the smaller size with high quality.
If you have a portable player, you almost certainly can't use Ogg's :)
If you make computer games, you have a high quality free way of adding a lot of music to your games. (possibly patents for mp3)
If you want background music in a computer game, why would you want to use a format that eats drastically more processing power?
You can do 44.1khz and 48 khz audio.
So can MP3, what's your point?
The encoder sounds good by default, so music traded on file sharing systems sounds good (unlike all those terrible 128k mp3s encoded by anything that isn't LAME).
So "The Encoder" for MP3 is bad? If there was just one encoder this would be an argument. And I do hat those 128k bastards just as much as you :) At least iTunes defaults to 160k.
Now the other points are very valid, but they probably won't get anyone to switch at this point. What we need is a format that gives at least 4x the compression of MP3 with the same quality (and reasonable CPU usage) to get people to switch. Hopefully it will be an open technology like Ogg.
"Reality is just a convenient measure of complexity" -Alvy Ray Smith
Ok, i'm just going to assume (based on this post by you and previous posts) that you work at microsoft.
One thing you should know is that i make embedded digital audio players for a living. I have been doing this for years. I have personally worked with every codec except mp3pro, and i doubt mp3pro will ever mature to market viability. i have seen and ported the wma decoder source, in addition to a variety of other minor things i could mention to provide credibility here.
* Much better than OGG and MP3
This is quantitative; most listening tests i have read about state that high quality mp3 encoders (such as lame) and the ogg reference encoder produce better quality output than WMA or AAC. I would guess that this trend will continue; Microsoft makes fast, low quality encoders for their desktop applications so as to provide an enhanced initial user experience. This is evidenced with how WMP behaves - it encodes as fast as possible, but generates low quality (notable artifacts) output, even at bitrates of 96kbps and 128kbps. This definitely refutes the claim that WMAv7 64kbps sounds "as good" as MP3 128kbps.
* Picture perfect at 128 kbit/s
No offense, but are you in the marketing department at MS? My response has to be "I'll believe it when i see it." I dont have the golden ears, but i can still tell 128kbps from cd audio, and i dont see this as changing.
* Supported by hardware (unlike ogg)
This is a flat out lie. Microsoft has ported their WMA decoder to various embedded architectures, but has no actual hardware support. The support is all in software, running on embedded processors. As was mentioned in previous posts, Ogg has been ported to embedded devices just like WMA; it's just a matter of time before it's ported to all devices.
* Next version (Corona) will sport 5.1 Dolby, 24 bit samples, 96khz sampling rate, better compression.
That's nice, except most consumer audio hardware handles 16 bit 44kHz audio, which is what CD audio is. So supporting 96kHz audio might look great on paper, but it does absolutely nothing for you in reality. In terms of 5.1 Dolby, AAC supports multiple channels and look where it's gone - nowhere. Maybe you guys should focus on the features that actually matter?
* Existing hardware will update firmware to support Corona
For the love of jesus. Let me drop you a clue:
* Existing hardware will update firmware to support OGG Vorbis 1.0
Your blind faith in WMAv8 has converted me - i am now a true believer in alternate technologies. I will devote all my spare time to the proliferation of disruptive technology.
Thank you for your support.
After collecting 60 Gb worth of mp3s, I switched to almost strictly shn format
over 2 years ago. Here is my reasoning:
1. Stick with a lossless format if you can afford the bandwidth and storage
space. Plan for the future, when bandwidth and hd space will be much
more plentiful.
2. I can definitely hear the difference between lossless and any compressed
format at 128 kb/s (that annoying wavery sound), and even at 256 kb/s (barely)
on very delicate passages and high-end speakers.
3. Also, if you want to reprocess the music (dehiss, dehum, equalize, normalize,
respatialize, etc) you experience a much more noticeable degradation in the
sound if you start with a lossy format.
4. shn is the standard format for trading music.
It is a lot less work to store in shn then have to decode and reencode every
time you make a music trade.
For lots of good links on shn format, see my trading page at
http://www.vsl.ist.ucf.edu/groups/vtb/TradeList
(Now that I've come this far, what the hell, trade requests here
.
;-)