vorbisPlus - OggVorbis + SBR emulation

[dimzon]

http://www.vlsi.fi/plusv/plusv.shtml

Test package: http://www.mytempdir.com/582387

Command line:
test.bat input.wav output.wav

Emulates OggVorbis + SBR @ ~48kbps

Unfortunally PlusV is avaluable under GPL licence and OggVorbis encoder/decoder is avaluable under BSD-like icence so them are incompatible...

I wrote to Xiph.org project manager asking to relicense Vorbis encoder/decoder library/SDK under dual license: BSD-like(current) and GPL to be able to extend and experiment with Ogg Vorbis using existing 3-td party GPL code without rewriting encoder from scratch.

Related threads @ HA forum:
http://www.hydrogenaudio.org/forums/index.php?showtopic=43277
http://www.hydrogenaudio.org/forums/index.php?showtopic=153

[Eastermeyer]

I think you need a Decoder supporting SBR , too ?

[dimzon]

I think you need a Decoder supporting SBR , too ?
Everything is in package
Actually it performs
Encoding input.wav->Decoding to output.wav at once - keep in mind it's just for testing to be able to evaluate how good is SBR for OggVorbis

[Eastermeyer]

I thought of an Winamp Plugin or something to Playback such Files.
But I understood , it's just for early Testing.

[GodofaGap]

@Dimzon: you are aware that you are in violation of the GPL and the Vorbis license by distributing these three binaries without these licenses, right?

Furthermore PlusV uses patented algorithms so it is way out of the philosophy of Vorbis.

[dimzon]

you are aware that you are in violation of the GPL and the Vorbis license by distributing these three binaries without these licenses, right?
No, I does not vioalate anything. I'm just using unmodified binaries of OggEnc and PlusV obtained from official sites (well, OggEnc gone from rarewares.org)

And keep in mind - this is just emulation to eveluate how sutable (in quality words) are using SBR technology for Vorbis

[shon3i]

Hmm, interesting, i now go to compare with CT,Nero,Mp3Pro

[GodofaGap]

No, I does not vioalate anything. I'm just using unmodified binaries of OggEnc and PlusV obtained from official sites (well, OggEnc gone from rarewares.org)
The GPL states that you MUST accompany binaries with the license (please read section 3). You are not doing this, so you are in violation. The fact that this is just for experimental purposes does not exempt you from this.

[dimzon]

Hmm, interesting, i now go to compare with CT,Nero,Mp3Pro
Please don't forget to share your results for us

@GodofaGap
OggEnc is NOT GPL. And there are not any license information in original archive dowloaded from rerewares.org
plusv has multiple licence sheme. And there are not any license information in original archive dowloaded from http://www.vlsi.fi/plusv/plusvdownload.shtml

Please stop blame me

[GodofaGap]

The binary distribution of PlusV is accompanied with the GPL so I don't know what you are talking about. Furthermore, the licensing conditions on this page (http://www.vlsi.fi/plusv/plusv.shtml#Using%20and%20licensing%20PlusV-technology) are quite clear. I don't know how you can have missed this.

[shon3i]

dimzon error

C:\vor>test.bat track.wav track2.wav

C:\vor>plusv.exe -e track.wav track.wav.a.wav track.wav.a.sbr
Encode
Input file is 16 linear bits @ 44100 Hz
Output file is 16 linear bits @ 22050 Hz
2 audio channels, PlusV bands 0..5 ~= bandwidth 11025..19293 Hz
Wrote 18.3 seconds (1402 blocks) at 5240 bits/s

C:\vor>oggenc2.exe -q 0 track.wav.a.wav -o track.wav.b.ogg
Opening with wav module: WAV file reader
Encoding "track.wav.a.wav" to
"track.wav.b.ogg"
at quality 0.00
[ 98.1%] [ 0m00s remaining] -

Done encoding file "track.wav.b.ogg"

File length: 0m 18.0s
Elapsed time: 0m 02.0s
Rate: 9.1579
Average bitrate: 43.5 kb/s


C:\vor>oggdec.exe track.wav.b.ogg


OggDec v1.9.2 (libVorbis 1.1.2) Compiled on: Dec 8 2005
Encoder Version: 0
Serial Number: 31412

Bitstream is 2 channel, 22050Hz
Scale = 1.0000
Decoded length: 403862 samples = 0:18 mins.
Encoded by: AO; aoTuV b4b [20051117] (based on Xiph.Org's libVorbis)
Decoding: track.wav.b.ogg

100% decoded.

********** Done decoding all input files. **********

C:\vor>plusv.exe -d track.wav.b.wav track2.wav track.wav.a.sbr
Decode
Input file is 16 linear bits @ 22050 Hz
Output file is 16 linear bits @ 44100 Hz
2 audio channels
Cannot read auxilliary data

C:\vor>test.bat track.wav track2.wav

C:\vor>plusv.exe -e track.wav track.wav.a.wav track.wav.a.sbr
Encode
Input file is 16 linear bits @ 44100 Hz
Output file is 16 linear bits @ 22050 Hz
2 audio channels, PlusV bands 0..5 ~= bandwidth 11025..19293 Hz
Wrote 18.3 seconds (1402 blocks) at 5240 bits/s

C:\vor>oggenc2.exe -q 0 track.wav.a.wav -o track.wav.b.ogg
Opening with wav module: WAV file reader
Encoding "track.wav.a.wav" to
"track.wav.b.ogg"
at quality 0.00
[ 98.1%] [ 0m00s remaining] -

Done encoding file "track.wav.b.ogg"

File length: 0m 18.0s
Elapsed time: 0m 02.0s
Rate: 9.1579
Average bitrate: 43.5 kb/s


C:\vor>oggdec.exe track.wav.b.ogg


OggDec v1.9.2 (libVorbis 1.1.2) Compiled on: Dec 8 2005
Encoder Version: 0
Serial Number: 32395

Bitstream is 2 channel, 22050Hz
Scale = 1.0000
Decoded length: 403862 samples = 0:18 mins.
Encoded by: AO; aoTuV b4b [20051117] (based on Xiph.Org's libVorbis)
Decoding: track.wav.b.ogg

100% decoded.

********** Done decoding all input files. **********

C:\vor>plusv.exe -d track.wav.b.wav track2.wav track.wav.a.sbr
Decode
Input file is 16 linear bits @ 22050 Hz
Output file is 16 linear bits @ 44100 Hz
2 audio channels
Cannot read auxilliary data

and when i try to see comand switches of plusv.exe says to type plusv.exe -h i type but don't show switches

[dimzon]

@shon3i
I have no idea.. Try longer sample
ADD
Seems like this error is for last chunk only - just ignore it

[shon3i]

@shon3i
I have no idea.. Try longer sample
I will

[shon3i]

But got same error. Maybe is not error becouse i got file somethig.wav which have size is like original wav but sound looks like sbr but is very terible

[dimzon]

@shon3i
Actually I discover this error affected last chunk only...
Anycase PlusV quality is terrible (maybe it's bug in plusV implementation, I don't know)
So we need better SBR encoder.
A asked Nero team some times ago but they don't answer
http://www.hydrogenaudio.org/forums/index.php?showtopic=42696&st=25&p=374008&#entry374008
Maybe we can write collective petition to Nero HE-AAC team about it (i believe it will not affect on their bisness)

[shon3i]

So we need better SBR encoder. Yes i agree with you but i think that Ivan & Graf can't make that app for free, becouse is wery powerfull

Maybe we can write collective petition to Nero HE-AAC team about it (i believe it will not affect on their bisness)
we can try

[dimzon]

but i think that Ivan & Graf can't make that app for free, becouse is wery powerfull
Why? Actually such application is not sutable for regular backup/encoding/streaming etc. It's sutable for testing/ABX puposes only (to look how good will be encoder X + SBR). Anycase they can restric this encoder by (for example) less than 5 min. sample duration etc...

I'm really VERY interested to listen @ Vorbis+SBR and, maybe, Vorbis+PS
And I want to ask Aud-X for same utility to test Vorbis+Spartial

ADD:
Maybe we can ask CT team for such tool too?

[shon3i]

Why? Actually such application is not sutable for regular backup/encoding/streaming etc. It's sutable for testing/ABX puposes only (to look how good will be encoder X + SBR). Anycase they can restric this encoder by (for example) less than 5 min. sample duration etc...I think will be only with restriction but will see, first to ask him

I'm really VERY interested to listen @ Vorbis+SBR and, maybe, Vorbis+PS
And I want to ask Aud-X for same utility to test Vorbis+Spartial
Aslo VorbisSBR will be great then i think that can do a new test @ 48kbs, aslo i wanna see that lame+SBR give same results like mp3PRO.

Maybe we can ask CT team for such tool too?
I don't know who is developer of CT, Is him shows on HA.

[shon3i]

Dimzon i can't belive did you try to encode to 48kbs with oggenc2.exe i get amaizing quality better than any HE-AAC codec, Vorbis is do a great job, aslo test multichannel is great.

[Ivan Dimkovic]

Tried it - and I don't find it so good (it is not bad, but not as good as HE-AAC IMO) - I guess I am much more sensitive to Vorbis' smearing/metallic sound and stereo field wobbling at such low bitrate, than to the "sandy" sound of HE-AAC.

Anyways, there will be a listening test soon - so please take part in it :) It will show a lot.

[gaekwad2]

Anycase PlusV quality is terrible
Indeed!
ABC/HR Version 1.0, 6 May 2004
Testname:

1R = C:\Programme\audiotools\vorbisPlus\Faron_Young-q0.wav
2L = C:\Programme\audiotools\vorbisPlus\Faron_Young_vorbisPlus.wav

---------------------------------------
General Comments:

---------------------------------------
1R File: C:\Programme\audiotools\vorbisPlus\Faron_Young-q0.wav
1R Rating: 4.5
1R Comment:
---------------------------------------
2L File: C:\Programme\audiotools\vorbisPlus\Faron_Young_vorbisPlus.wav
2L Rating: 1.5
2L Comment:
---------------------------------------
ABX Results:
C:\Programme\audiotools\vorbisPlus\Faron_Young-q0.wav vs C:\Programme\audiotools\vorbisPlus\Faron_Young_vorbisPlus.wav
16 out of 16, pval < 0.001
(4.5 is too high, but compared with the 'vorbisPlus' file regular q0 sounded almost transparent)

[shon3i]

Tried it - and I don't find it so good (it is not bad, but not as good as HE-AAC IMO) Vorbis uses VBR so maybe this explanation why is so good, I am now listen to my stereo speakers but when i found my hedaphones i then say it is good or not.

sensitive to Vorbis' smearing/metallic sound and stereo field wobbling at such low bitrateI think i hear that and other artifacts but i'am not sure. Tomorow i post results.

@gaekwad2 yes but real ogg vorbis is not so good with PlusV, this "vorbis 2" aotuv is actualy optimized for low rates.

[gameplaya15143]

since shon3i mentioned that they still haven't fixed the -h (help) option, and if dimzon didn't remove the hard-coded encode delay from plusv.exe, than it WILL sound very bad because the plusv stream will not be in sync with the wav file. the reference plusv.exe still has the delay hard-coded in it (the delay is there because mp3 is not natively gapless)

I have fixed the -h option, and removed the 32khz input samplerate limit (you can use 16khz as input if you like), and added a switch to change the plusv high-band. Take a look at the plusv documentation if you don't know the meaning of what I just said.

http://www.geocities.com/gameplaya_15143/downloads
Now you can do proper testing... well almost, there is still a 1 sample delay, but that shouldn't be enough to notice. (I could be wrong on the delay units, but it's 1 of something)

(i couldn't test what dimzon did, tempdir wouldn't load)

------------------
This is my first post, so hello everyone. I've been a reader of doom9 for like 3 years now :thanks:

[gaekwad2]

I'm not sure that helped much, anything with snare drums or cymbals still sounds horrible.

@gaekwad2 yes but real ogg vorbis is not so good with PlusV, this "vorbis 2" aotuv is actualy optimized for low rates.
Sorry, I don't quite understand what you mean, I used the same oggenc at the same quality setting (the one from dimzon's zip, which is based on aoTuVb4.51).

[gameplaya15143]

as far as proof/disproof of concept goes...

plusv.exe -e %1 %1.a.wav %1.a.pv
oggenc2.exe -q -2 --advanced-encode-option lowpass_frequency=99 %1.a.wav -o %1.b.ogg
oggdec.exe %1.b.ogg
plusv.exe -d %1.b.wav %2 %1.a.pv
try that instead, bitrate should average closer to 32kbps (use the plusv.exe from my previous post, otherwise you have a delay of 700some samples or so between the bottom half and the top half of the audio spectrum)

note: the lowpass must be effectively off, otherwise you end up with a gap in the audio spectrum, which is not a good thing

and if you really like fun... (only use my plusv.exe build)

plusv.exe -e %1 %1.half.wav %1.full.pv
plusv.exe -e %1 %1.quarter.wav %1.half.pv
oggenc2.exe -q -2 --advanced-encode-option lowpass_frequency=99 %1.quarter.wav -o %1.quarter.ogg
oggdec.exe %1.quarter.ogg
plusv.exe -d %1.quarter.wav %2.half.wav %1.half.pv
plusv.exe -d %2.half.wav %2 %1.full.pv
run:
test.bat input.wav output.wav
use 44100hz audio as input, the vorbis part will be 11025hz (could be done 3 or 4 times with 48khz input) :cool:
(i didnt actually test it in a batch file, i just did it manually.. i'm sure anyone reading this can handle it)

i forgot to mention this before... plusV is NOTHING like sbr.
I think plusV has a LOT of potential, but after reading the specs.. I think it's being held back. I would like to see it without any 'base' codec