I'm trying to create some listenable files at a low bitrate so I decided to fool around with Ogg Vorbis...MPC just isn't going to give me good results @ ~100kbps.

Over the last two days, I've encoded 7 albums using -q2.5. I've been rather impressed with the quality at this bitrate (nominal=104kbps). While you aren't fooled into thinking the output is transparent to the original, the overall sound is pleasing and not distracting or annoying. If I am to complain, the ogg files have a certain fuzzy or veiled demeanor about them and there seems to be less air and spaciousness in the treble. But there is little MP3-like swishiness and smearing to be heard. It's nice to encode an hour of music in under 50MB.

I've used Ogg RC3 encoders in the past but don't remember them sounding as good on low bitrates as the one I recently downloaded from Rare Wares. Have there been compiled improvements to the code since RC3 was released in 12/2001?

try mppenc --thumb and aacenc -radio and come back

Oggenc -q2.5 is certainly better than mppenc --thumb (though I haven't tried the latest 1.02 version) and I'm too impatient to deal with AAC.

I'm not abandoning MPC altogether...I just have a low bitrate need for a certain "project" of mine that isn't suited to the format.

I have also tried few cross-comparisons with MPC 1.02 --thumb, aacenc 2.15 -radio and oggenc RC3 -q 1.5.

Needless to say, all of these encoders introduce serious distortion. It's about choosing one with the least displeasing artifacts, and Vorbis does it for me. MPC encoded file has this tremolo effect whereas hihats and cymbals sound incrediably weird with aacenc. Somehow Vorbis manages to sound surprisingly balanced and it's artifacts are least annoying.

In this particular clip aacenc IMO sound much worse than MPC --thumb. And just listen how oggenc dominates .

http://www.cs.uku.fi/~tnkorhon/Tristania-ruins-intro.flac

I tried it, and I find AAC to be best sounding and Ogg worst. MPC is in the middle. I was suprised to see MPC beat Ogg, especially since the bitrate was somewhat lower.

Hmm. Matter of taste seemed to surface really quickly on this one .

I condemned aacenc mainly because of the way hihats sounded beginning at 0:06. They sounded extremely repulsive to me. Vorbis had pretty much hissing and noise but I liked that much better than the aacenc way.

Edit: typo

Latest build of aacenc also has a -qvbr X switch, to control VBR by quality directly.

X could range between 0 and 100 - maybe this setting would work well.

AACEnc does not use "fake" (Intensity) stereo - and because of this Ogg sounds more "bright" at very low bitrates, for the cost of worse stereo image.

@mithrandir

To answer you earlier question, YES, there have been many major lib revisions since RC3 and they are incorporated in the version you downloaded from RareWares. If the lib version appearing in the encoded file says 20020408, you have the latest real version.

john33

To my knowledge there aren't yet low bitrate optimizations implemented in post-RC3 lib.

I could be completely wrong on this, and I've never used CVS but I hear that Monty's branch in the CVS of Vorbis could have some lbr optimisations. However I did say it could have - not that it has!

Ruairi (from a dark alley near you)

Yes, this is the lib version I have. So if you are right about the revisions to the post-RC3 lib, I am not imagining these improvements.

I'm impressed by even -q0. There's the occasional nasty artifact, but it's the best 64kbps performance I've heard from an encoder using stereo and bandwidth exceeding 10KHz.

The post-rc3 CVS compile at rarewares does not include new low bitrate optimizations like "noise normalization" or new masking data.

It's mostly code cleanup and faster pre-echo control/short block triggering control.

I'm actually using the OggDropXPd tool, which says it uses "standard" Ogg libraries.

oggdropXPd is simply a GUI frontend to the standard libvorbis. It contains many extensions to the official oggdrop.