Encoders/decoders built using LAME 3.97 beta 1 2005-09-29

Bundle: includes lame.exe, lame_enc.dll (ICL4.5) and lameACM.acm (ICL9.0)
Download (594Kb)

lame.exe fixed for .mp2 and mp3 decoding: lame.exe only (ICL4.5) - also includes amended source code files and diff files. For mp3 and mp2 decoding, this now yields the same output as foobar2000 but the error checking remains unchanged.
2005-10-15
Download (222Kb)

The difference is pretty much as indicated. The fixed version has mp2 decoding encabled and, more importantly, the mp3 decoding takes account of the encoder and decoder delays so that the output is sample accurate compared to the original source.

So, if you use lame for decoding, yes, you may wish to use this version. If you don't, it is of no interest to you.

pretty much any other method of decoding mp{2,3} would be better than using lame...


later

pretty much any other method of decoding mp{2,3} would be better than using lame...

Yes. People should be using Nitrane or DigiDeck.

In the wiki page doesn't appear these decoders ? Do you know why?
DigiDeck is recommended too ?
The development stopped in mid-1998 and in the wiki page of HA says:
Older decoders could provide bad quality, due mostly to encoder implementation errors. Such bad encoders include Winamp before version 2.666 (using Playmedia's low quality AMP engine), decoders based on old XAudio, and Digideck. Also, some older decoders did not work well with VBR streams.

Retrieved from "http://wiki.hydrogenaudio.org/index.php?title=Best_MP3_Decoder"

Gr.
Gonzalo

Those words are sarcastic. He did not really recommend you to use Nitrane or DigiDeck.

1. I was NOT sarcastic.
2. Anyway, rjamorim is the only one who can says if I was sarcastic or not, not you.

Gonzalo

EDIT : I figured this out http://www.hydrogenaudio.org/forums/index....topic=30820&hl=

i believe he meant that rjamorim was being sarcastic, not you.    =o

Nop, he did not mean that.

OK, just to clear this thing.
I use LAME 3.97b1 to encode to MP3.
And I also use it to decode to WAV.
Should I continue using LAME for decoding ?
Should I use another decoder ?

Gr.
Gonzalo

There are a couple of reasons I decode using Foobar's diskwriter instead of Winamp's. These won't necessarily matter for people's mp3 files...
1. Foobar reads LAME headers to do gapless decodes. Whereas Winamp will add 576 samples of basically silence, at the beginning, and whatever number of leftover samples at the end of the track. This only really matters with live music, or songs that are supposed to be gapless.
2. With foobar you can dither the output. This is useful if you have significantly lowered the volume of the mp3 files using replaygain - it's possible that you'll get a noise floor that's too low with the outputted wave files. And this doesn't matter for files that haven't had their volume lowered with replaygain, or have only been dropped a small bit.

Mp3 files are in frames of 1152 samples each. CD-audio tracks are in "frames" (basically) of 588 samples (75 of these per second, for a total of 44100 samples per second).
Most mp3 encoders introduce a delay of some samples into the first frame before the actual info from the encoded wav begins. In LAME, this delay is 576 samples.
Also note that since mp3 files need to be an integer of frames (you can't really have half a frame, or if you can, your mp3 header information is messed up), there will be some null samples at the end.
LAME deals with this by writing info to the header on the number of samples of delay (at the start) and padding (at the end). So a file might say
enc_delay: 576
enc_padding: 1644
Foobar reads this information and does not play (or decode) the initial 576 samples, or the last 1644, thus providing true gapless playback. So you end up with a decoded file the exact length of the initial encoded wav file.
Winamp ignores this information, and thus plays (and decodes) the extra samples. And it's not always clean silence, sometimes with a very slight baseline noise, so you can't actually get rid of the initial 576 samples from a winamp decode by "removing null samples".

So the upshot: if decoding a live album that was encoded with LAME, use foobar.

LAME uses mpglib, which is known to be reliable, but doesn't decode gaplessly.
foobar2000 uses mpglib too, decodes gaplessly and lets you use its advanced dithering algorithms for optimal quality beyond audibility.

I would not trust Lame's decoding too much.
There are probably still some uncorrected bugs in the mpglib version we are using.

Foobar2000 uses a custom modified version of mpglib as dev0 already said and should be absolutely safe for decoding LAME encoded mp3 to wav. Plus you will be sure that all your tracks will be decoded gapless.