Rationale for LAME Development
by Mike Cheng, September 1998
A few years back I started hearing about this fantastic new sound compression technology; MPEG layer 3. I had used layer 2 before and was very impressed, and so I started to look for layer 3 software.
The Amiga
At this time I was a dedicated Amiga user, and I went about finding source code to port over to the Amiga. I spent much time optimizing the dist10 decoder so that I could at least play back these files. Then AMP came along and I gave up optimizing and just ported AMP to the Amiga. Then Stephane Tavenard came along and made a superfast m68 asm mp3 player for the Amiga and I bowed to his greater programming prowess. However, there was an encoder in the dist10 source as well, and I set about porting this to the Amiga.
Now the Amiga is a good machine with a fab OS, but it just doesn't have the computational grunt to do MP3 encoding at a reasonable speed. I spent lots of 1997 just playing with the dist10 source code, optimizing little bits and making it faster. This optimizing effort appeared as an amiga-only program.
Changing Times
Early this year a group called 8hz announced that they were going to start a concentrated effort to speed up the ISO source code. I immediately signed up. Within a week I had sped up the already fast 8hz effort by incorporating my speedups into their code. Over the next few months, I'd improved on the 8hz source code - speeding it up by about 45%. All these changes I made public and mailed back to the 8hz guys. It was good to be part of this net effort.
Fraunhofer Crackdown
In September 1998, Fraunhofer started cracking down on use of their freely available source code (don't ask me why), and the 8hz effort, and a number of other free encoders based on ISO sources were also halted. That sucked.
Now..
Now I've released a patch file to the mpeg ISO source code. And this is called LAME which is gnu-speak for LAME Ain't an MP3 Encoder. [For non-english speakers, "ain't" is a way of saying "is not"]. I haven't released an MP3 encoder. I've released a patch file for the dist10 source code. The file I am releasing is totally incapable of producing an MPEG layer 3 file. The file I am releasing cannot be compiled, and is not an executable. You must use the ISO MPEG source code in order to get a functional MPEG layer 3 encoder.
Open Source
The SlashDot community has got it right. Open source is the way to go. BladeEnc's closed source annoys me, because the author is not sharing his ideas. [Ed. Note: BladeEnc is now under the LGPL] Due to licensing arrangements I doubt he can make money on BladeEnc, so why not release the source? The claim that it will lead to versions with inferior quality is bogus. If you leave yourself as the authoritative source for the code, it doesn't matter what everyone else does with it - so why not release it? It can only lead to better software...
To the future...
I've been thinking about an open source community effort to develop a compressed audio format equivalent/better than MPEG layer 3. In thinking about it, I simplified it down to make it seem like an achievable task:
- Only 44.1khz samples. This is CD quality. Don't bother with 32 or 48khz.
- Only 128kbit/s bit rate. The mpeg standard tries to cover bitrates from 16kbit/s up to 320. That's a huge range and requires different acoustic and computational models. 128kbit/s produces CD-quality sound for MPEG layer 3, AAC and VQF.
Anyone who has ideas on this is welcome to get in contact with me.
