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: Speex beta 4 is out! (Read 7246 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Speex beta 4 is out!

I've just released Speex beta4, which will hopefully be the last beta before starting release candidates. There are a number of quality improvements and a some new features. First, the quality of variable bit-rate (VBR) has been greatly improved, mostly for wideband. Also, some improvements were made to the packet-loss concealment code in order to reduce artifacts caused by lost packets. There has been improvements to the 2.15 kbps mode, which now sounds a bit better. The last quality improvement is in the perceptual enhancement which is now good enough to be enabled by default in speexdec (but not by default in libspeex for now).

One of the new features in this release is a VAD (Voice Activity Detection) mode that allows to save bit-rate when there's no speech even without VBR enabled. This VAD comes with DTX (Discontinuous Transmission) that allow to completely stop transmitting when the noise is constant (in which case, comfort noise generation is turned on). The last new feature in this release is an ABR (Average Bit-Rate) option, which uses VBR with a quality that's adjusted to meet a certain target (average) bit-rate.

Also, I've finally updated the samples for beta4. Checkout Speex at 4 kbps VBR!

Speex beta 4 is out!

Reply #1
Win32 binaries.


Speex beta 4 is out!

Reply #3
any winamp plugins available for speex?


Speex beta 4 is out!

Reply #5
I'm experimenting with Speex for the first time and would like to know if the command lines I am using need any tweaking for better results.

I recorded myself saying what the male said in your examples and made 3 seperate mono versions of the original... 32khz, 16khz, and 8khz.  All done in cooledit pro 2 with the highest quality downsampling.

(for 32khz) speexenc -u --quality 0 --vbr --dtx --comp 0 test1.wav test1.spx
(for 16khz) speexenc -w --quality 0 --vbr --dtx --comp 0 test2.wav test2.spx
(for 8khz) speexenc -n --quality 0 --vbr --dtx --comp 0 test3.wav test3.spx

[here is a zip file with everything in it]

Speex beta 4 is out!

Reply #6
Quote
I'm experimenting with Speex for the first time and would like to know if the command lines I am using need any tweaking for better results.

(for 32khz) speexenc -u --quality 0 --vbr --dtx --comp 0 test1.wav test1.spx
(for 16khz) speexenc -w --quality 0 --vbr --dtx --comp 0 test2.wav test2.spx
(for 8khz) speexenc -n --quality 0 --vbr --dtx --comp 0 test3.wav test3.spx

Well, I hope you're not expecting good quality with --quality 0 (the quality setting goes from 0 to 10). Also, the --comp option goes from 1-10 and changes the amount of CPU the encoder is allowed to use. By increasing the complexity (--comp), you can (slightly) increase quality at the same bit-rate (at the cost of more CPU). Two other things to note: DTX is implied by --vbr and the -n/-w/-u options are not necessary for wav files because the sampling rate is taken from the wav file. None of these is an error though.

Speex beta 4 is out!

Reply #7
OH  I thought lower numbers where better quality  I'm used to LAME switches  thanks.

speexenc --quality 10 --vbr --comp 10 test1.wav test1.spx

Is that good?

I have a 2.52ghz, so I dont think a high --comp leval will hurt, but we'll see.  maybe it will for long ass files.

[edit] well i must say, that sure sounds a whole lot better.  And may I also say, wow, this is truely a great compression format you have made!  It really puts all other encoders to shame at these small bitrates with vioce

I will get much use out of it.  THANKS

Speex beta 4 is out!

Reply #8
hello Imvalin, merry Christmas

as it looks like speexdrop V004 includes speexenc v1.02 (according to winamp file info) i am in search for the downmix to mono and resample to lower Khz
(44100 to 16000) options within the command line speexenc v1.04. I can't  find anything like this reading -help.
Thanks in advance

Speex beta 4 is out!

Reply #9
Quote
hello Imvalin, merry Christmas

as it looks like speexdrop V004 includes speexenc v1.02 (according to winamp file info) i am in search for the downmix to mono and resample to lower Khz
(44100 to 16000) options within the command line speexenc v1.04. I can't  find anything like this reading -help.
Thanks in advance

I will update speexdrop over the next few days. I would have done it before, but it needs some functionality added to cater for the new options.

Speex beta 4 is out!

Reply #10
John33 you are the best,

many many thanks for all your efforts. I like very much all your *drops, i use them all day.

As about speexdrop004, i thought that it made use of the external speexenc -dec. Now that i realised that it's an all in one package i have a question. What resampling routines do you use and (quality wise) are there any benefits if i first resample my wavs with ssrc-hp and then use the resampled ones in speexdrop? (using olny the downmixing if needed).

Speex beta 4 is out!

Reply #11
Quote
Quote
hello Imvalin, merry Christmas

as it looks like speexdrop V004 includes speexenc v1.02 (according to winamp file info) i am in search for the downmix to mono and resample to lower Khz
(44100 to 16000) options within the command line speexenc v1.04. I can't  find anything like this reading -help.
Thanks in advance

I will update speexdrop over the next few days. I would have done it before, but it needs some functionality added to cater for the new options.

This sounds great john33. I can say that the original sources for most material I use is 44KHz and would like to see resampling done at encode time. Of course, the source files I am referring to are losslessly stored (APE) but never mind that for the moment.

Is SSRC being considered for downsampling? Truth is I don't know how effective SSRC is outside 44->48 KHz conversion.
"Something bothering you, Mister Spock?"

Speex beta 4 is out!

Reply #12
Speexdrop currently uses the resampling routines out of oggenc!! These are also the same ones used in oggdropXPd. I was considering looking at using ssrc instead, but I haven't got round to it yet!!  However, I wouldn't have thought that considering the target use of speex, the use of ssrc was particularly critical?

 

Speex beta 4 is out!

Reply #13
Ok, I thought it was only OggDrop that had resampling.
I mentioned SSRC since it boasts both speed and accuracy.

Anyway, my error thinking resampling did not exist yet. You are probably right about SSRC being extravagant for Speex low bitrates. Thanks for your reply john33.
"Something bothering you, Mister Spock?"

Speex beta 4 is out!

Reply #14
Quote
I will update speexdrop over the next few days. I would have done it before, but it needs some functionality added to cater for the new options.

Sorry for the delay, but speexdrop V0.0.6 (using beta 4 libs) is now uploaded to and available from my 'Others' page at Mirror 1.