Aften 0.0.8 is out

[jruggle]

jruggle:

Is psychoacoustic implementation in your recent todo list?

Yes. I actually implemented a very simple one, but it's only preliminary and not ready for inclusion yet.

[tebasuna51]

What is the right command line to encode 6 mono wav to AC3 5.1 with aften?
The help from aften.exe say nothing for 6 mono wav.

One more method to encode 6 mono wav to ac3 5.1:

CopyAudio -I "" -S "FL FR FC LF BL BR" fl.wav fr.wav fc.wav lf.wav bl.wav br.wav - | aften -readtoeof 1 - output.ac3

Where CopyAudio.exe can be obtained from AFsp-v8r2.tar.gz at Audio File Programs and Routines (http://www-mmsp.ece.mcgill.ca/Documents/Software/index.html)

-I "" is to avoid extrachunks at end of file

-S "FL FR FC LF BL BR" force WAVE_FORMAT_EXTENSIBLE header with specified channelmask. Not necessary because is aften default.

Tested with equivalent multichannel wav > 4GB.

[LigH]

BeSweet is also able to multiplex 6 mono WAV files to a 6-channel WAV. You can use BeLight as GUI, it contains a MUX file wizard. Select the preset "WAV" (or another which is the same - I think AAC) for the channel order.

I am just not sure if BeSweet can handle resulting WAV files >4 GB correctly.

...


I do remember that support for MUX files in Aften was requested for some time already.

[DarkAvenger]

I do remember that support for MUX files in Aften was requested for some time already.

Well, the source is open. Anybody can step forward with a patch.

[tebasuna51]

BeSweet is also able to multiplex 6 mono WAV files to a 6-channel WAV. You can use BeLight as GUI, it contains a MUX file wizard. Select the preset "WAV" (or another which is the same - I think AAC) for the channel order.
Yes, I say other after the Kurtnoise post (http://forum.doom9.org/showthread.php?p=1026242#post1026242) about BeSweet and the mine (http://forum.doom9.org/showthread.php?p=1026311#post1026311) about AviSynth methods.

I put this last method because don't require extra files like .mux or .avs, and can be easy implemented in GUI's

I am just not sure if BeSweet can handle resulting WAV files >4 GB correctly
Also have problems with > 2GB.

[canuckerfan]

okay... i've got a 5.1 AC3 file here which is pretty messed up. basically I want to convert it to mp3. but here's the problem... its gots some serious noise issues and all 5 channels (except of course the LFE) sound EXACTLY the same - and my ears are pretty sensitive. this is hinting me that the source was originally mono and after doing a little research i've confirmed that the source was mono and was smoshed up into a 5.1 mix. so this is the workflow that i've proposed...

1) convert the ac3 file into a 5.1 wav file. then demux one of either R, L, C, RR, LR.
2) remove noise/etc with goldwave
3) convert single wav mono file into mono mp3.

but my problem is what happens to the LFE channel? does that data get lost when I go to mp3 since I'm only working with one of the channels? any input is appreciated.

[DarkAvenger]

@canuckerfan

Did you read the topic before hijacking the thread?

[canuckerfan]

^i did read the topic. others have done the same so I thought it'd be ok. i'm sorry, my intentions weren't to hijack this thread. I will post a new thread.

[Nikos]

Thanks again tebasuna51 for the usefully suggestions.

One more question, in aften i must use -pad 0 or -pad 1?

[tebasuna51]

One more question, in aften i must use -pad 0 or -pad 1?

-pad 1 is the default -> 5.33 ms delayed like SoftEncode and others.

-pad 0 -> without delay, but with a 'fade in' in first 5.33 ms.

[Nikos]

I want to convert 6 mono wavs from a DTS-HD file to AC3 with correct Dialog Normalization value.

The whole center channel containing dialogue, sound effects and music give me RMS level -22 db with Sound Forge.

The average RMS level with Sound Forge at several places with dialogue and very little music or sound effects was -31 db.

The Sound Forge settings in the normalize window was Ignore below: -45 db and Use equal loudness contour.

Which is the correct value for the Dialog Normalization in aften?

[planet1]

http://kurtnoise.free.fr/index.php?dir=Aften/&file=aften_rev547.zip

Could the pipe be broken :confused: :eek: :confused:

[tebasuna51]

http://kurtnoise.free.fr/index.php?dir=Aften/&file=aften_rev547.zip

Could the pipe be broken :confused: :eek: :confused:

And also command line don't run in my XP sp1.

[Kurtnoise]

http://kurtnoise.free.fr/index.php?dir=Aften/&file=aften_rev547.zip

Could the pipe be broken :confused: :eek: :confused:
you have to force the raw switches for the pipeline....just like :

C:\>ffmpeg.exe -i "E:\Music\Rock el Casbah.wav" -f wav - | "C:\temp\aften.exe" -raw_fmt s16_le -raw_sr 44100 -raw_ch 2 -b 448 - "E:\Music\FFm_aften222.ac3"



@tebasuna : any error message ? if you have a multicore, try to force the threads #...

[tebasuna51]

@tebasuna : any error message ? if you have a multicore, try to force the threads #...

>aften wavex.wav z.ac3

Aften: A/52 audio encoder
Version SVN - r547
(c) 2006-2007 Justin Ruggles, Prakash Punnoor, et al.

input format: WAVE Signed 16-bit little-endian 48000 Hz 5.1-channel
output format: 48000 Hz 3/2 + LFE

SIMD usage: MMX SSE SSE2 SSE3
Threads: 1

and abort. The same forcing the -threads 1
Testing with XP sp1 in Pentium 4

Maybe SSE3?

edit: Work with Pentium Dual Core and Vista

[planet1]

you have to force the raw switches for the pipeline.

Thx for the answer, for the moment I'll use Wisodev's builds with foobar2000.

[DarkAvenger]

I'd try using latest svn. I fixed some bugs in the cpu detection code which could lead to crashes.

[tebasuna51]

@Kurtnoise
Your aften_rev531 and aften_rev552 work from 0% to 100% send the 'Done!' message but never end in pipe method (Bepipe or Wavi), after a Ctrl+C to finish don't exist output file:

>bepipe --script "NicDtsSource(^blade.dts^)" | aften - zzz.ac3

Aften: A/52 audio encoder
Version SVN - r552
(c) 2006-2007 Justin Ruggles, Prakash Punnoor, et al.

***************************************
BePipe by dimzon
***************************************
Script used:
# BEGIN
NicDtsSource("G:\Pelis\Ptes\blade.dts")
# END


Scanning for Audio Stream...
Found Audio Stream
Channels=6, BitsPerSample=16, SampleRate=48000Hz
Writing Header...
Writing Data...
Done!
^C
>

The last working rev for me is aften_rev521.

[Kurtnoise]

As I said earlier, now we need to force the raw switches for the pipeline...

bepipe --script "NicDtsSource(^blade.dts^)" | aften -raw_fmt s16_le -raw_sr 48000 -raw_ch 6 - zzz.ac3

[tebasuna51]

As I said earlier, now we need to force the raw switches for the pipeline...

Is true, I don't read the precedent post completely, only my part. Sorry.

But, what is the problem now. A new regression?

Aften always read the headers before, if all is considered raw data the headers are converted also to sound with channel async. problems.

>bepipe --script "WavSource(^z6.wav^)" | aften -raw_fmt s16_le -raw_sr 48000 -raw_ch 6 - zz6.ac3

Aften: A/52 audio encoder
Version SVN - r552
(c) 2006-2007 Justin Ruggles, Prakash Punnoor, et al.

***************************************
BePipe by dimzon
***************************************
Script used:
# BEGIN
WavSource("z6.wav")
# END


Scanning for Audio Stream...
Found Audio Stream
Channels=6, BitsPerSample=16, SampleRate=48000Hz
Writing Header...
Writing Data...
0% input format: RAW Signed 16-bit little-endian 48000 Hz 6-channel
6-channel audio must have LFE channel
error initializing encoder
Done!

>

A perfect wav 6 chann are rejected because the 44 headers bytes is not a multiple of 12 (number of bytes per 6 chan sample).

A stereo 16 bit are converted (44 = 11 x 4) with initial click.
A stereo 32 bit (8 x ? = 44) are also rejected.

Aften can´t work now with actual BeHappy, Bepipe or Wavi.

[Kurtnoise]

But, what is the problem now. A new regression?
Dunno why actually... The only thing which might be the problem I guess, it's the raw pcm input support introducing in the revision 529.

I'll try to upload several builds tonight for testings. but I'm pretty sure it's not me that broke the code because I haven't change a lot of thing in it...:p

[tebasuna51]

I'll try to upload several builds tonight for testings. but I'm pretty sure it's not me that broke the code because I haven't change a lot of thing in it...:p

Well, I don't think you are the culprit. :)

But is easy to code, if -raw_fmt is not present the header must be read.

[DarkAvenger]

I found the bug, I will post it on the aften ml...

OK, I commited a fix - I hope so.

It is obviously wrong to activate raw mode when piping a wav. This works for me (now) in Linux:

cat in.wav |./aften - out.ac3

[Kurtnoise]

Ok...rev 661 is up for some tests.

It is obviously wrong to activate raw mode when piping a wav.
I thought that when the source had some different parameters from the default settings, raw switches were mandatory. So, I was wrong...


btw, I've a question: my PC supports as well 3DNow! & 3DNow!ext. However, these SIMD are not displayed in the frontend during encoding (only MMX...SSE3). What's wrong ?

[tebasuna51]

@Kurtnoise13, DarkAvenger
Seems rev561 work now.

Thanks

[DarkAvenger]

btw, I've a question: my PC supports as well 3DNow! & 3DNow!ext. However, these SIMD are not displayed in the frontend during encoding (only MMX...SSE3). What's wrong ?

While the cpu detection routine does detect those SIMD instruction sets, it won't get printed out, as Aften doesn't contain such code, yet.

[wisodev]

I just uploaded my R561 build (http://sourceforge.net/project/showfiles.php?group_id=183195&package_id=232924) of Aften (aften-svn-r561-vs2005.zip). This is VS2005 (but only soulution and project files used, not compiler) build but due the latest changes in Aften I was forced to use Intel C++ Compiler to be able to build Aften correctly (missing headers * pmmintrin.h * and error connected to missing headers) in Visual Studio 2005 SP1. There are no Win64 binaries included and no static libraries.

Thanks,
wisodev

[DarkAvenger]

...build but due the latest changes in Aften I was forced to use Intel C++ Compiler to be able to build Aften correctly (missing headers * pmmintrin.h * and error connected to missing headers) in Visual Studio 2005 SP1.

I don't think I changed anything there. I rather have the impression your defines are borked. Using cmake it detects that vs2005 doesn't support SSE3 and thus won't try to include pmmintrin.h. As you don't use cmake, you have to take care on your own.

BTW, Orcas has SSE3 support and I have succesfully build it with nmake generator.

[Yobbo]

Excuse please, I'm a bit confused with Aften versions!? :confused: I am using WAVtoAC3Encoder 0.4 with Aften A/52 v0.06 courtesy of Wieslaw! But I see there is Aften v0.07 now? Is there a gui for 0.07? What is the latest up-to-date? Can somebody be kind and please enough to tell me how to keep up-to-date with Aften plus gui?

Thank you! :)

[Kurtnoise]

WavtoAC3Encoder or AftenGUI both work with the last official release (0.07).

and you can use also the lastest Aften revisions with them (links are above).

[Yobbo]

well I got WavtoAC3Encoder here, which says it uses Aften 0.06. I just downloaded Wisodev's 0.07 R561 build from a few posts up... Now what do I do? Sorry to be a pain :scared:

[Kurtnoise]

Just put it the GUI folder...

[Yobbo]

What exactly do I put in the gui's folder? The whole Aften folder (there's lots of stuff in there!). Or, just the Aften.exe which is in the Win32 subfolder? And also, where exactly do I pop it? In the Gui's main folder? Or in the Win32 subfolder? Or the "ansi" sub-subfolder? Sorry again for the newb questions!

[Yobbo]

No one can help me to update my Aften gui? :(

[Kurtnoise]

If you use AftenGUI 1.4 and Aften.exe revision 561, you're up to date.

Then just put aften.exe in the same AftenGUI folder, load your files , choose a target bitrate and go !

[Yobbo]

OK thanks!! :)

[DarkNite]

I just wanted to drop in and say thank you to everybody working on Aften and AftenGUI. I appreciate it.

:thanks:

[mltan]

Hi! Just want to clarify this: " I have substituted the FFmpeg MDCT implementation with the one from libvorbis, which is faster. " - Justin @ Aftenblog

Because I found out that the one from libvorbis is based on "The use of multirate filter banks for coding of high quality digital audio" by Sporer et al. But this paper was presented a very long time ago (1992).

I am wondering why this has been chosen over other implementations and why there has been no updates since version 0.06 on this. If I am not mistaken, I think there are other more efficient methods, right?

How does it compare against other mdct implementation in terms of speed and quality, why is it "faster" then?

@wisodev and kurtnoise, i would just like to ask what you used to create the GUIs. (I am planning to create one that will operate stereo settings only)

Thank you very much!

[Kurtnoise]

mine's developped in Delphi. I would like to create a X-plateform GUI (gtk+ or something with the Aften API) but I 've a lack of time.

[wisodev]

@wisodev and kurtnoise, i would just like to ask what you used to create the GUIs. (I am planning to create one that will operate stereo settings only)

Thank you very much!

My GUI is created in Visual Studio 2005 using Visual C++. The app is based on MFC Dialog template. You can download latest source code of my app from Subversion repository (http://sourceforge.net/svn/?group_id=158644) (here you can browse online the source code (http://thefrontend.svn.sourceforge.net/viewvc/thefrontend/EncWAVtoAC3/)) or just by downloading source package (http://sourceforge.net/project/showfiles.php?group_id=158644&package_id=219726).

wisodev

[mltan]

bout the GUIs: wow, that was fast! I'll probably check out the MSVC first since its the one im using now. to kurtnoise and wisodev, :thanks:

[sl1pkn07]

sl1pkn07@SpinFlo:~/aplicaciones$ svn co https://aften.svn.sourceforge.net/svnroot/aften aften-0.07-svn
........
Revisión obtenida: 561
sl1pkn07@SpinFlo:~/aplicaciones$ cd aften-0.07-svn/
sl1pkn07@SpinFlo:~/aplicaciones/aften-0.07-svn$ cmake .
-- Check for working C compiler: /usr/bin/gcc
-- Check for working C compiler: /usr/bin/gcc -- works
-- Check size of void*
-- Check size of void* - done
Please do an out-of-tree build:
rm -f CMakeCache.txt; mkdir -p default; cd default; cmake ..; make
CMake Error: in-tree-build detected
-- Configuring done
sl1pkn07@SpinFlo:~/aplicaciones/aften-0.07-svn$

use cmake version 2.4-patch 6 in Kubuntu X86_64

[Kurtnoise]

Please do an out-of-tree build:
rm -f CMakeCache.txt; mkdir -p default; cd default; cmake ..; make
it's clearly mentioned...

[sl1pkn07]

VERY LOL!

sorry ><

im supposed that script worked for me :S

[jruggle]

I am wondering why this has been chosen over other implementations and why there has been no updates since version 0.06 on this. If I am not mistaken, I think there are other more efficient methods, right?

How does it compare against other mdct implementation in terms of speed and quality, why is it "faster" then?

The main reason for using the libvorbis implementation was that the original MDCT was from FFmpeg, which was (and is still) not very fast. In fact, there is a discussion going on now at ffmpeg-devel about creating a new FFT (and hence MDCT) implementation.

Also, cpu-optimized versions of the libvorbis mdct were already out there ready to use. This wasn't part of the decision, but has been a definite advantage.

I'm sure there are faster implementations out there. At one time DarkAvenger was doing some experiments using FFTW with Aften. It was a while ago, so I don't remember if it was any faster.

[jruggle]

I just released Aften 0.0.8. Yes, I changed the version numbering. I hope it's not too confusing. I think the new versioning scheme will be easier to use. Anyway, here is the Changelog.


fixed piped input from FFmpeg
added support for MPEG channel order remapping
restructured audio input. enables raw pcm file support.
bugfixes in MMX/SSE2 code
stack align hack for x86 MinGW with threads
API changes
SIMD and threads usage is shown and is configurable
screen output gets updated every 200ms to reduce load
SIMD detection changed to compiler-independent inline assembly, thus nasm/yasm not needed anymore

[patul]

@wisodev: Any chance you would update your Wav to AC3 Encoder with this new release? :D

[wisodev]

@wisodev: Any chance you would update your Wav to AC3 Encoder with this new release? :D

Yep, I have just released new version of WAV to AC3 Encoder (http://www.thefrontend.net/EncWAVtoAC3/index.html) at version 0.5. You can download it from here (http://www.thefrontend.net/EncWAVtoAC3/index.html) (changelog and downloads archive (http://www.thefrontend.net/EncWAVtoAC3/download.html)).

Thanks,
wisodev

[madshi]

I just released Aften 0.0.8. Yes, I changed the version numbering. I hope it's not too confusing. I think the new versioning scheme will be easier to use. Anyway, here is the Changelog.


fixed piped input from FFmpeg
added support for MPEG channel order remapping
restructured audio input. enables raw pcm file support.
bugfixes in MMX/SSE2 code
stack align hack for x86 MinGW with threads
API changes
SIMD and threads usage is shown and is configurable
screen output gets updated every 200ms to reduce load
SIMD detection changed to compiler-independent inline assembly, thus nasm/yasm not needed anymore

Thank you. I appreciate especially the raw pcm file support. One question about this: Do you expect the same channel order in the raw pcm file as you do in the wav file? Or do you expect the Blu-Ray pcm channel order?

[jruggle]

Thank you. I appreciate especially the raw pcm file support. One question about this: Do you expect the same channel order in the raw pcm file as you do in the wav file? Or do you expect the Blu-Ray pcm channel order?
Same channel order as WAVE unless you specify the option to change it. What channel order is Blu-Ray?