What do we know about the new NCTU-LAME encoder? Is it a serious competitor to LAME 3.90.3 or is it just optimized for speed like GOGO?

http://dsppc14.csie.nctu.edu.tw/projects/i.../nctu-lame.html

All I know is that the NCTU-FAAC outputs worse quality than the current Faac build (1.20.1)
For instance, there's no block switching (all blocks are long).

More info by Ivan:
http://www.audiocoding.com/phorum/read.php...1&i=3721&t=3717

That sounds like SBR...

The codec does implement it. I'd like to know whether they're better off than PlusV.

These guys are obviously breaking the LGPL (Lame's license), by not posting sources with their binaries.

This looks interesting. There will have to be some tests done to prove that it's better quality than lame is at right now.

The fact that a Prof. and 3 Dr.'s are working on it gives more hope.

I hope that there are more releases and more improvements to this!

dj_digital

Edit: I just tried the encoder and it looks like it's based on lame 3.88. My first encode went at 1.68x max.(compared to 3.56x in lame 3.90.3 on same song). So far, I think that lame 3.90.3 has the edge but I haven't experimented with the command options yet. First beta too so it has room to grow....(hopefully)

Well.. even if they could improve many areas of psychoacoustics, it's still far away from the max-tweaked setup.
I'm not expecting NCTU professors and doctors to tweak it to the max, rather just introduce some proof of a concept psychoacoustic which might or might not be better. It's a long way to perform better in practise than 3.90.3 --alt-presets...

If they couldn't even top FAAC 1.20.1 with their NCTU-FAAC (shouldn't really be the hardest job), I'm not expecting much from their Lame version either.

hehehe, i know what you mean..

Well.. I tested it a bit. I'd almost say (the proprietary high freq encoding excluded) the biggest difference comes from their recommended --noshort, which increases spectral resolution in many cases. It's not exactly so bad even though their recommended setting doesn't use short blocks at all.

If you use this --noshort, it will still retain full mp3 compatibility right? Don't know much about the tech of mp3 so could you clear this up for me?

Thanks,

dj_digital

--noshort is still mp3 compatible

The SBR needs the side information encoder to enhance the quality. The HFR is free from the constraint. The technology try to reconstruct the high frequency contents to have the better quality without the help of the encoder information. It tries to enhance the usual audio files.


There are constraints from the hybrid filterbank, the long window is not that efficient. The objective of an encoder is the high quality. The tools used is not so important. Compatibility is the only issue.


Whether or not the FAAC is better than the NCTU-FAAC is not related to the quality of Lame. The tools of Lame is much more restricted. We need to find the constrained optimum solution.


We hope it is so easy too. It will let us loss the interstes of challenges too.