Comment 63 for bug 195483

Revision history for this message
KennoVO (kenno-xs4all) wrote :

- This is not about bit rate, this is about quality. Getting a higher bit rate does not automatically equal getting a higher quality. In fact, botching the psychoacoustics model with a load of settings is a good recipe to get similar or higher bit rates but a lower quality.
- For assessing quality, one person's subjective listening tests mean next to nothing. It is for this reason that I kept my own perception of sound quality out of this discussion (I also have a set of pretty good headphones and trained ears, but that's simply not relevant). However, there are ways to objectively test quality, by using ABX methodology, long systematic testing, a large sample of listeners, and sounds statistical analysis. That 's exactly what the folks at hydrogenaudio.org did, and they came up with recommended settings for standalone LAME. They also found that it's easy to pull down the quality by giving LAME a bunch of extra switches (you'd have to go digging through the hydrogenaudio forums for that). For this reason, I will trust anything that gives output that is bitwise identical to standalone LAME with the hydrogenaudio recommended switches, and I will not trust anything else. The "lamemp3enc" element falls into the former category, the "lame" element into the latter. Histograms of bitrates are irrelevant.
- Conicidentally, the same people at hydrogenaudio.org found that "Nowadays LAME is considered the best MP3 encoder at mid-high bitrates and features the best VBR model among MP3 implementations." Oh, and there's what they think about Xing: "The Xing (pronounced "zing") MP3 encoder was one of the fastest MP3 encoders. Of course that comes with a price, and quality wasn't on par with encoders tuned for quality instead of speed, like FhG Slowenc (Audioactive) and LAME."

So here we are, LAME is at the very least one of the best encoders out there, and nobody will miss the restricted ones that were thrown out. As FiloSottile said, the "lame" interface to liblame is buggy, the "lamemp3enc" interface is not. The only thing up for discussion is really how to call "lamemp3enc" (and make sure it's implemented in the Maverick release). Here's my proposal again:
https://bugzilla.gnome.org/show_bug.cgi?id=630779

CD Quality
----------
audio/x-raw-int,rate=44100,channels=2 ! lamemp3enc name=enc target=0 quality=2 ! xingmux ! id3v2mux

Portable MP3 Player Quality
---------------------------
audio/x-raw-int,rate=44100,channels=2 ! lamemp3enc name=enc target=0 quality=6 ! xingmux ! id3v2mux

Voice Quality
-------------
audio/x-raw-int,rate=44100,channels=2 ! lamemp3enc name=enc target=1 bitrate=56 mono=true ! xingmux ! id3v2mux