Help - Search - Members - Calendar
Full Version: Best Lame Settings For Phone-quality Audio?
Hydrogenaudio Forums > Lossy Audio Compression > MP3 > MP3 - Tech
Bluenote
Hi... I am trying to encode a series of recordings off a phone answering machine using LAME 3.92. The audio is currently in WAV files, not too high quality: 22kHz, 8 bit, mono. Are there any recommended settings for this type of audio? I couldn't find anything appropriate in the LAME presets, but I somehow doubt --alt-preset standard is what I should be using. wink.gif

Thanks in advance for any tips!
Hanky
It depends on how important the size of the encoded files is to you.
I would throw in something like --alt-preset 50 -a --lowpass 7 --resample 22
If you are not limited to the mp3 format, you could also try to use Ogg Vorbis at quality -1, which gives bitrates in the 40-50 kbps range and much better quality.
/\/ephaestous
If you don't care about using a format other than mp3, you could use Speex, It's designed to ncode voice and I'ts part of the Vorbis proyect.
I haven't used it thougth, so I can't say anything about its quality
jmvalin
QUOTE(/\/ephaestous @ Nov 11 2002 - 07:23 PM)
If you don't care about using a format other than mp3, you could use Speex, It's designed to ncode voice and I'ts part of the Vorbis proyect.
I haven't used it thougth, so I can't say anything about its quality

If you're going to try Speex, I suggest down-sampling to 16 kHz (though it's not absolutely necessary) and then you can get good quality using VBR at around 20 kbps. BTW, Speex is part of Xiph.org which also has Vorbis, but it's not a "Vorbis project" as Vorbis is a different codec. But codecs use Ogg though.
M
Bluenote, you might want to check this other thread in the forums; JohnV did a considerable amount of testing on such parameters, and posted his results there. Briefly, here are the settings he came up with, as well as a link to L.A.M.E. 3.94 alpha2 (which, according to JohnV, gives better results with these lines).

24kbps speech:
--alt-preset cbr 24 -a --resample 22 --lowpass 7 -Z

16kbps speech:
-b 16 -a --resample 11 --lowpass 5 --athtype 2 -X3

8kbps garbage (NOT recommended; described as sounding "ugly"):
-b 8 -a --resample 11 --lowpass 4.0

Hope this is helpful - and give a round of thanks to the person who did all the testing.

- M.
Hanky
Please read the original Q,
Bluenote does not explicitly ask for ultralow bitrates, he even considered using plain --aps.
M
QUOTE(Hanky @ Nov 12 2002 - 09:41 AM)
Please read the original Q,
Bluenote does not explicitly ask for ultralow bitrates, he even considered using plain --aps.

I did read the original question; Bluenote asked for settings which would handle low-quality audio, and for the purposes described (archiving messages from an answering machine) the low bitrates seemed advisable. They still do.

- M.
Bluenote
Thanks guys, for all your help.
Speex sounds promising but I think for now I'll stick to a codec I already have on my machine...
OGG would be possible, but I think I'll have a bash at some of the LAME settings M so helpfully listed. Might wait till 3.94's out of alpha though...

Thanks again!
jmvalin
QUOTE(Bluenote @ Nov 10 2002 - 08:35 AM)
Hi... I am trying to encode a series of recordings off a phone answering machine using LAME 3.92. The audio is currently in WAV files, not too high quality: 22kHz, 8 bit, mono.

Just noticed you're using 22 kHz for phone quality. This is a waste since the phone is limited to ~3500 Hz, so down-sampling to 8 kHz would be a good idea (it won't remove any information at all). Also, at 8 kHz using Speex you could get transparent quality around 15 kbps or average-good quality at ~6 kbps in VBR mode.
JohnV
QUOTE(M @ Nov 12 2002 - 07:48 AM)
Bluenote, you might want to check this other thread in the forums; JohnV did a considerable amount of testing on such parameters, and posted his results there. Briefly, here are the settings he came up with, as well as a link to L.A.M.E. 3.94 alpha2 (which, according to JohnV, gives better results with these lines).

24kbps speech:
--alt-preset cbr 24 -a --resample 22 --lowpass 7 -Z

16kbps speech:
-b 16 -a --resample 11 --lowpass 5 --athtype 2 -X3

8kbps garbage (NOT recommended; described as sounding "ugly"):
-b 8 -a --resample 11 --lowpass 4.0

Well, like I said in that thread, it's better to use ABR unless CBR is mandatory:
24kbps speech:
--alt-preset 24 -a --resample 22 --lowpass 7 -Z
16kbps speech:
--abr 16 -a --resample 11 --lowpass 5 --athtype 2 -X3

The CBR had to be used for special (flash) purposes by the original question maker.
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.
Invision Power Board © 2001-2008 Invision Power Services, Inc.