Hello all, I'm currently trying to encode some mp3 tracks for a class, and because of the nature of these tracks about 60-70% of it is dead air. I am currently using --alt-standard, but the problem is there is just enough noise during the dead air that it is encoding the silence at 128k instead of 32k. Can anyone suggest some settings to raise the "Dead Air" threshold so the silence is encoded at 32k?


Here is a link so you can get a sense of what I'm talking about

You could try alt-medium instead, though I'm not sure that's going to be truly helpful. Beyond that I'm not sure. Someone more knowledgable can hopefully provide an answer as I'm curious too.

If it was me, I'd run the WAV through some noise reduction filters before encoding. Dump the background noise where I could do so. Clean it up a bit.

I was going to mention that originally, but when I listened to the example posted it didn't seem like there was very much noise in there to clean up. Definitely worth trying though.

You should try adding -b32 at the end of your parameters. IIRC, --aps uses a minimum of 128kbps regardless of content.

So try --alt-present standard -b32

Thanks for the comments, I was thinking about noise reduction but that ment having to redo a TON of ID3 tags/filenames by hand since I'm using EAC for this. Wondering, why not set -b32 for --alt-standard in general? Shouldn't the aglorithm be smart enough to work without this hard floor at 128K?

The presets use 32 kbps when digital silence is detected - if no signal is above the ATH

You could raise the ATH for your recordings. But be aware that if you exaggerate it this will produce nasty artifacts, almost like a primitive noise gate.

Just an update. Setting the floor to 32k dropped the average bitrate from ~128 to ~90, which ment I could fit all I needed to one one CD which was my goal. I'm still curious about the hard floor at 128k for -aps though, I am wondering have any tests been done without the floor?

How about splitting the wav in parts and applying seperate settings to each part
and then joining it to one file again?

For my personal use of aps with music i'm always setting '-b 112' particularly on tracks where left and right channel seems to be very similar.

You get - in that special case - a big profit by setting a low minimum for your special tracks with many silent parts. But please remember, aps is tuned for music and music is something like an opposite of silence (ahh well, ...this doesn't mean, music always has to be loud!).

I must admit, i don't remeber clear reasons why it is set at 128kbps and me too i have some problems to explan this.
Maybe there are so few complaints in the user's hope to get 'safe' files. Because also for a soft/low passage of stereo music neither 112 nor 128kbps is much. Moreover it makes quite no difference for the total filesize of a couple of tracks of modern music if '-b 32' or '-b 128' is set. Make a test yourself on some music tracks - if you save over 100kB on one single track encoded with aps i would call it much.

During 3.96 beta testing, the lower limit for aps was set to 96 I believe. Among some of the samples, which sounded inferior to 3.90.3, were a few which could be fixed simply by raising the minimum bitrate back to 128. The psychoacoustic model sometimes makes these mistakes.

With most music you don't see many < 128kbps frames anyway so space savings would be minimal.

IIRC, if you use lame 3.96.1, --aps will now work thru <128 bitrates.