Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Want to match quality of WMA Audio 9 Voice at low bitrate (Read 6167 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Want to match quality of WMA Audio 9 Voice at low bitrate

Don't even know if I'm in the right place but am working on a project to load 100's of hours of spoken word onto portable mp3/wma players.  We have used the WMA 9 Voice codec at 16kbps and really like the quality of the sound considering the small file size, but the players are not compatible with that codec for play back.  And I seem to be reading about this incompatibility everywhere on the web.
I have been playing around with Lame3.98 and all2lame front end to try to get a good quality sound at 16kbps ABR but cannot come close to the clean sound we get from the WMA codec at the same rate.
This has probably been talked to death by many of you but sending this new guy to the right place would be very much appreciated.
Thanks in advance.

Want to match quality of WMA Audio 9 Voice at low bitrate

Reply #1
I have been playing around with Lame3.98 and all2lame front end to try to get a good quality sound at 16kbps ABR but cannot come close to the clean sound we get from the WMA codec at the same rate.


Voice codecs work much better for voice then do general audio codecs.  I don't know of any portable players that can do WMA Voice, although maybe some smartphones can.  Rockbox devices can do speex which is similar, and may eventually do WMA Voice.

Otherwise AAC+ or WMA std may be worth trying, as some portable devices do support them, although they won't do as good.

Want to match quality of WMA Audio 9 Voice at low bitrate

Reply #2
I have tried the regular WMA, it does play on the player but you are right that it does not sound as good as the WMA Voice or the mp3's I've been experimenting with. 
Thanks for your speedy response.  Still hoping for some string of switches that will come close.
And BTW why is there not a voice version of an mp3 codec? Seems there would lots of call for such.
Thanks again.

Want to match quality of WMA Audio 9 Voice at low bitrate

Reply #3
And BTW why is there not a voice version of an mp3 codec? Seems there would lots of call for such.


They're fundamentally different technologies.  The equivalent of MP3 for voice encoding would probably be the CELP family of encoders, of which WMA Voice is one variety.

Want to match quality of WMA Audio 9 Voice at low bitrate

Reply #4
We have used the WMA 9 Voice codec at 16kbps and really like the quality of the sound considering the small file size, but the players are not compatible with that codec for play back.

It's unfortunate but the bottom line is that with a prerequisite for player compatibility, you're essentially stuck with mp3.  And due to the inherent limitations of its technology you're not going to get anything decent at around 16 kbps.  So unless space is a significant factor, you might want to accept that for compatibility using mp3 at a higher bitrate is necessary.

There's a quasi-official voice setting for mp3: --abr 56 -mm (56 kbps ABR, mono; see here.).  Perhaps you could try this at 48, 32, or 24 kbps and see if it yields acceptable results.

 

Want to match quality of WMA Audio 9 Voice at low bitrate

Reply #5
For me, WMA voice @16kbps is somewhere between lame --abr 24 -mm  and lame --abr 32 -mm.

Want to match quality of WMA Audio 9 Voice at low bitrate

Reply #6
Space is the determining factor for us as well as availability in a low-cost mass-market consumer MP3 player.
We would like to develop a custom player in the future with a voice codec, but for now; what would be the optimum set of parameters for LAME for maximum capacity without creating something that is uncomfortable to listen to for long periods.
I think we're getting pretty good results at 16 ABR with both a high and low pass filter but still get an annoying flanging/ringing sound (my description) that would be great to get rid of.
You have been hugely helpful and I appreciate the input.
Thanks again.

Want to match quality of WMA Audio 9 Voice at low bitrate

Reply #7
Try the following with lame (if you need stereo, remove the mono flag):

Code: [Select]
-V8 -m m --resample 24

I'm always happy with the filesize/quality ratio.

Want to match quality of WMA Audio 9 Voice at low bitrate

Reply #8
I'd suggest GXlame which is optimized for low bitrate, if you choose mp3. If AAC is acceptable then Nero CLI is probably best of all.

Want to match quality of WMA Audio 9 Voice at low bitrate

Reply #9
If you have control over the players (it sounds like it), buy compatible models (e.g., Sansa Clip) and just install Rockbox on all of them.

Rockbox does  not yet support WMA Voice but it supports Speex, which will give you very low rates at high quality.

You can even fully customize Rockbox to suit your needs.


Want to match quality of WMA Audio 9 Voice at low bitrate

Reply #10
I'd suggest GXlame which is optimized for low bitrate...


Try GXLame -V0 -mm --resample 24

You'll get somewhere between 16-32kbps for speech (usually around 20kbps). Quality should be quite good. (If you want lower, use --resample 16 instead. Then you'll get around 16kbps or less.)
Note that this also works well for mixed content (general microphone recordings, movie audio, etc).
Copy Restriction, Annulment, & Protection = C.R.A.P. -Supacon

Want to match quality of WMA Audio 9 Voice at low bitrate

Reply #11
Thanks for the GXLame tip.
Seems to work better than the lame.exe for low (I'm shooting for 16 or 22) bit rates. Still not as good as the windows audio 9 voice but getting there.
I've notices a huge speed difference if I use the -h switch.  I'm batch encoding nearly 500 hours of voice (about 800 files) so speed helps but is not a necessity. 
Is there a big quality difference with or without the -h switch.  I can't hear one but want to ask the question before I begin.
Thanks again for all of your help.

Want to match quality of WMA Audio 9 Voice at low bitrate

Reply #12
Thanks for the GXLame tip.
Seems to work better than the lame.exe for low (I'm shooting for 16 or 22) bit rates. Still not as good as the windows audio 9 voice but getting there.
I've notices a huge speed difference if I use the -h switch.  I'm batch encoding nearly 500 hours of voice (about 800 files) so speed helps but is not a necessity. 
Is there a big quality difference with or without the -h switch.  I can't hear one but want to ask the question before I begin.
Thanks again for all of your help.


The difference is not tremendous, no. It's often inaudible or barely perceptible, but the overall bitrate very slightly lower. Also, GXLame is intended to target quality levels, not specific bitrates. For that reason, something like "GXLame -V0 -mm --resample 24" is quite a bit better than specifying a bitrate of 22kbps, for instance. To hit nearer 16kbps, try "GXLame -V0 -mm --resample 16." I believe this will yield the same sampling frequency as that of WMA Voice at comparable bitrates. I find the quality slightly more accurate than WMA Voice (as there are fewer psychoaccoustic 'tricks' used to single out voice material). For this reason, it's quite a bit better on mixed (voice and other sound/music) than WMA Voice, in my experience, and certainly a fair bit more compatible.

Apologies for not working much on GXLame after t4 -- I targeted the release for a specific low-bitrate audio test that was scheduled to be conducted on this forum but was later canceled due to various reasons. Without such a test looming, my motivation to work on the codec is limited.
Copy Restriction, Annulment, & Protection = C.R.A.P. -Supacon