Aften 0.0.8 is out

[tebasuna51]

Thanks wisodev.

There are some channel coupling improvement in last Aften version?

[jruggle]

Thanks wisodev.

There are some channel coupling improvement in last Aften version?
There really have not been many changes in a while. Just some tweaks to the exponent strategy decision. There were a few code structure changes in preparation for channel coupling though. And it looks like Prakash has added Windows x64 support while I was away for Christmas. :)

[TFM_TheMask]

New version of WAV to AC3 Encoder is available for download , version 4.3 includes libaften.dll version from 20091226 (Git master head snapshot).
http://code.google.com/p/wavtoac3encoder/downloads/list

Isn't it possible to only get 2 aften libraries which can detect the PC's SIMD capabilities (SSE, SSE2, SSE3, MMX). So only libaften_x86.dll and libaften_x64.dll and not 6 or more different libraries?

So what I mean is a compiled libaften library with all SIMD instructions in it and it checks which instructions it can use depending on a PC's capabilities.

[wisodev]

Isn't it possible to only get 2 aften libraries which can detect the PC's SIMD capabilities (SSE, SSE2, SSE3, MMX). So only libaften_x86.dll and libaften_x64.dll and not 6 or more different libraries?

So what I mean is a compiled libaften library with all SIMD instructions in it and it checks which instructions it can use depending on a PC's capabilities.

All my libaften.dll builds have enabled ALL Aften SIMD optimizations (if your hardware supports specific SIMD it is enabled on runtime). But additionally the MMX, SSE, SSE2, etc. builds have extra optimization made by Intel C++ Compiler and REQUIRE presence of SIMD support in hardware (else you can expect crushes), so you just can use default builds: libaftendll_x86 and libaftendll_AMD64 most of the time.

[TFM_TheMask]

All my libaften.dll builds have enabled ALL Aften SIMD optimizations (if your hardware supports specific SIMD it is enabled on runtime). But additionally the MMX, SSE, SSE2, etc. builds have extra optimization made by Intel C++ Compiler and REQUIRE presence of SIMD support in hardware (else you can expect crushes), so you just can use default builds: libaftendll_x86 and libaftendll_AMD64 most of the time.

Ok thanks, will do that. One little question, the libaftendll_AMD64 is only for AMD x64 processors or also Intel x64 processors?

[wisodev]

Ok thanks, will do that. One little question, the libaftendll_AMD64 is only for AMD x64 processors or also Intel x64 processors?

It's AMD and Intel CPU compatible (actually it's compiled using Intel C++ Compiler).

[TFM_TheMask]

It's AMD and Intel CPU compatible (actually it's compiled using Intel C++ Compiler).

Better change the name to x64 instead of AMD64.

:thanks:

[raquete]

New version of WAV to AC3 Encoder is available for download , version 4.3 includes libaften.dll version from 20091226 (Git master head snapshot).
http://code.google.com/p/wavtoac3encoder/downloads/list
thank you so much wisodev!
(no more 'old problems' running Win2000, changing to Win7)

veru very Happy New year! :)

[MrVideo]

Better change the name to x64 instead of AMD64.

Ya, had me confused. I downloaded the AMD64 versions, only to discover that it is for 64bit OSs. I have the AMD64 dual core, but run 32bit OS.

[MrVideo]

I've been using an older version of the files found in the aften_x86 directory.

Which is better, to use the files in the aften_x86 directory, or the files in the libaftendll_x86 directory?

If the libaftendll_x86 contents are used, I'm going to assume that they are placed in the same directory as I've previously done with the aften_x86 contents.

There was nothing in the top level readme file about which should be used, and why.

Thanks.

[canuckerfan]

does encwavtoac3 support stereo wav input or do I need to demux the stereo wav into 2 mono wav files?

[Boulder]

No need for that, it supports stereo WAVs just fine.

[canuckerfan]

but when I hit encode it gives me an error saying "Supported are minimum 2 and maximum 6 mono input files!"

[Boulder]

I just tried encoding a 32-bit stereo WAV with the latest version, no problems here.

[canuckerfan]

^maybe its my file... I tried the first file on this webpage ( M1F1-Alaw-AFsp.wav): http://www-mmsp.ece.mcgill.ca/Documents/AudioFormats/WAVE/Samples.html

and I got the same error. does that one work for you?

ps: the second one also gives me the same error(M1F1-AlawWE-AFsp.wav). somethings definately wrong with my setup. i am running v4.4 from the zipped package

[Boulder]

I think it's a problem with the encoding type in the WAV file.. try some "regular", i.e. PCM encoded WAV. For example this one is such: http://www-mmsp.ece.mcgill.ca/Documents/AudioFormats/WAVE/Samples/AFsp/M1F1-int16-AFsp.wav

[wisodev]

does encwavtoac3 support stereo wav input or do I need to demux the stereo wav into 2 mono wav files?

When you are using "Mmultiple mono input" mode the input files must be mono, if you turn off this mode then you can encode single multichannel input file.

[canuckerfan]

I think it's a problem with the encoding type in the WAV file.. try some "regular", i.e. PCM encoded WAV. For example this one is such: http://www-mmsp.ece.mcgill.ca/Documents/AudioFormats/WAVE/Samples/AFsp/M1F1-int16-AFsp.wav
I get the same error with that file as well.
When you are using "Mmultiple mono input" mode the input files must be mono, if you turn off this mode then you can encode single multichannel input file.
how do I disable mono input mode?

[tebasuna51]

Uncheck the 'Multiple mono input' box.

[canuckerfan]

Uncheck the 'Multiple mono input' box.
wow. I can't believe I missed something like that:eek:

thanks

edit: I have another problem now... my wav source is not sampled to 48k and I have to feed an avs script into encwavtoac3 so that it will convert it. here's my script:

#avisource("G:\temp\1.avi")+avisource("G:\temp\2.avi")
#SoundOut()

WavSource("G:\temp\bgf.wav")
ssrc(48000)

however as soon as I drop the avs into the program it waits about a second and gives an error. "Failed to initialize avisynth". then I have to end task it. I tried playing the avs file in mpc and works just fine... btw i'm running AviSynth 2.5.8

edit: I just found out that it only works with avs 2.5.7 and below as of Aug 2/'09. I guess that's still the case. I'll try it on 2.5.7 and give it a whirl.

[b66pak]

@canuckerfan or you can forget about drag&drop and use the add button...
_

[canuckerfan]

^I can't believe that made a difference haha. thanks, it works now.

[wisodev]

@canuckerfan or you can forget about drag&drop and use the add button...
_

Should be fixed in version 4.5 :)

[Maxiuca]

I keep getting "Encoder Error: Failed to initialize encoder". Funny thing is that when I start the program for the first time after system restart I can get it to encode, but then this error keeps appearing in logs. I'm on 64-bit Vista Ultimate. Any ideas?

[wisodev]

I keep getting "Encoder Error: Failed to initialize encoder". Funny thing is that when I start the program for the first time after system restart I can get it to encode, but then this error keeps appearing in logs. I'm on 64-bit Vista Ultimate. Any ideas?

What is the input file format ? Are you using 64 or 32 bit version ? Is the program configuration preserved after restart ?

[Maxiuca]

Sorry for a false alarm, problem solved, my bad. I was trying to encode a 5.0 ac3 with LFE low-pass filter enabled. It was pure coincidence that the program started encoding after two reboots (different settings).

[Midzuki]

Let's say, I have a 3min 36sec stereo WAV file (extracted from an Audio CD), and compress it with Aften.exe @ 192kbps, and then use spdifer.exe to obtain a DDWAV stream. Instead of getting a 36.3MB .wav, spdifer produces a 34.8MB one (and if I try a different bitrate, e.g., 256kbps, the resulting SPDIFed .wav has only 17.4MB :eek: ). This problem *does not* happen with ac3 files generated by Soft Encode, for example.

[Midzuki]

OK, today I found this post by madshi:

http://forum.doom9.org/showpost.php?p=1231929&postcount=945

which made me give a try to the latest release of EncWAVToAC3. Sadly the "44.1kHz problem" seems to be even worse in the libaften-GUI encoder: the same .WAV source-file mentioned in my post above becomes a 5.77MB .ac3 encoded at 224kbps by EncWAVToAC3, however spdifer.exe obtains a
12-byte .WAV file from that. :eek: Again, spdifer keeps working correctly on files created by a "Dolby-certified" encoder...

[tebasuna51]

OK, today I found this post by madshi:

http://forum.doom9.org/showpost.php?p=1231929&postcount=945

The problem don't exist now.

BTW, I can't understand the process.

You have a stereo wav file 3m:36s (36,3 MB), encoded with ac3 192 Kb/s (4.9 MB) and spdifer convert the file to ddwav (34.7 or 36.3 MB)
What is the benefit to have a lossy compressed audio with the same size than the lossless source?

With my stereo samples, encoded with Aften and Soft Encode, obtain the same wrong output size when use spdifer.

I have some 5.1 ddwav and the size isn't the size of the wav sources. Is the size of the ac3 x 2.2
ac3 file 31.5 M -> ddwav file 69.6 MB

With spdifer I obtain the same ratio with 5.1 ac3 files.
Maybe spdifer don't work properly with stereo files.

[Midzuki]

It depends on the spdifer's version one uses, I think. :confused: FWIW, I'm using the build 0.1a.

Anyway, I am not thinking in terms of "benefits", just in terms of "standard-compliance" --- if someone intends to create a multichannel AC3 Audio CD, it's better that they know they shouldn't be using Aften for doing that. :devil: Given that spdifer does work correctly with the 32kHz && 48kHz AC3 outputs from Aften, it seems to me, *now*, that the 44.1kHz encodes from Aften are not "correct enough". I remember that "many months ago", just for testing/learning purposes,
I created some stereo AC3s @ 44.1kHz with aften.exe, and realized that most of them were played back @ 48kHz when decoded by AC3Filter. Because such files were played back @ 44.1kHz when decoded by ffdshow, I (erroneously) concluded that the actual problem "had to be" in AC3Filter, "not" on the 44.1kHz files generated by Aften.exe.

P.S.: So, ¿¿¿ is there any better alternative to spdifer.exe ???

P.P.S.: ffdshow (r1376) audio processor is not an option,
because its AC3 encoder supports only 48kHz and 32kHz.

[tebasuna51]

...
--- if someone intends to create a multichannel AC3 Audio CD, it's better that they know they shouldn't be using Aften for doing that.
For what?
With spdifer.exe (ac3filter_tools_0_31b) I obtain the same filesize, framesizes and headers when convert to ddwav a 5.1 wav encoded to ac3 with Aften or Soft Encoder.

P.S.: So, ¿¿¿ is there any better alternative to spdifer.exe ???

I don't know.

[Midzuki]

With spdifer.exe (ac3filter_tools_0_31b) I obtain the same filesize, framesizes and headers when convert to ddwav a 5.1 wav encoded to ac3 with Aften or Soft Encoder.

I have some 5.1 ddwav and the size isn't the size of the wav sources. Is the size of the ac3 x 2.2

Something is wrong, then.

Maybe spdifer don't work properly with stereo files.

It DOES work with every stereo .ac3 @ 32 and 48kHz I have tested so far.

What are you trying to say, man? :)

P.S.: Another test --- stereo WAV @ 44.1kHz, encoded with AC3 ACM via VirtualDub; resulting file was "RIFF-stripped", and then given to spdifer.exe;
DDWAV file size = 36.3MB, as it should be. :rolleyes: :devil:
----------------------------------------------
UPDATE:

Just downloaded and tested the latest version of spdifer, and
this time, it did produce a DDWAV with correct-length from an Aften 44.1kHz file :) :) :)

So,

1) the real problem was in the old version of spdifer.exe, OK ;

2) BIG QUESTION: what would be the "magic" differences between the 44.1kHz outputs from Aften and the ones from the other encoders ??? :confused:

[jruggle]

2) BIG QUESTION: what would be the "magic" differences between the 44.1kHz outputs from Aften and the ones from the other encoders ??? :confused:
My first guess is that some other encoders do not vary the frame size. The frame size can be the base frame size or +16 bits at 44.1kHz per the AC-3 spec. Aften uses this option to make sure the average bitrate is exact. Otherwise it would be slightly less than the nominal bitrate. This is not an issue for 48kHz or 32kHz because the math is nice and even. :) Maybe the old version of spdifer did not take this into account?

If you want a reference, see document ATSC A/52B Table 5.18.

[Midzuki]

My first guess is that some other encoders do not vary the frame size. The frame size can be the base frame size or +16 bits at 44.1kHz per the AC-3 spec. Aften uses this option to make sure the average bitrate is exact. Otherwise it would be slightly less than the nominal bitrate. This is not an issue for 48kHz or 32kHz because the math is nice and even. :) Maybe the old version of spdifer did not take this into account?

If you want a reference, see document ATSC A/52B Table 5.18.

Thanks for answering. :)

Yes, I had already looked at the "top-secret" .PDF. :D
But, as you can see, it appears it was NOT ONLY spdifer.exe that didn't consider the «better» possibility. I mean, it seems Aften was/is the first/only "well-known" AC3 encoder to use two different frame sizes for the 44.1kHz sampling rate.

Again, :thanks:

[tebasuna51]

... I mean, it seems Aften was/is the first/only "well-known" AC3 encoder to use two different frame sizes for the 44.1kHz sampling rate.

Nope, the test I make before encoding the same wav 44.1KHz file:

Aften: 3161 frames (2536 "short" + 625 "long"), precise bitrate
Soft Encode: 3161 frames (1585 "short" + 1576 "long"), imprecise bitrate

[Midzuki]

Nope, the test I make before encoding the same wav 44.1KHz file:

Aften: 3161 frames (2536 "short" + 625 "long"), precise bitrate
Soft Encode: 3161 frames (1585 "short" + 1576 "long"), imprecise bitrate

:thanks: for the accurate info.

*THUMBS UP*

[Selur]

I wanted to add support for aften inside Hybrid but ran into the problem that aften doesn't show progress when I pipe into it from ffmpeg.
Here are the command lines I use:
ffmpeg -v -10 -threads 4 -i "input.ac3" -acodec pcm_s16le -ac 6 -ar 48000 -f wav - | aften -b 256 -readtoeof 1 - "output.ac3"
and
ffmpeg -threads 4 -v -10 -y -i "input.ac3" -f u16le -acodec pcm_s16le - | sox --temp "D:\Encoding Temp" -t raw -e signed-integer -2 -c6 -r48000 - -t wav - remix -m 1v0.2646,3v0.1870,4v0.1870,5v0.2291,6v0.1323 2v0.2646,3v0.1870,4v0.1870,5v-0.1323,6v-0.2291 norm | aften -b 128 -readtoeof 1 - "output.ac3"
I know that when feeding aften directly with a file progress works, but since I want to be able to filter with sox and not use intermediate files I really would like too stay with the piping the inptu into aften.

I know that '-v' shows me the current frame processed, but I don't know if I can somehow judge from the frame number to the current position processed.

So if someone knows how to get aften to show progress (e.g. through altering my command line) or knows how to conclude from the current frame to the current position any help is welcome.

Cu Selur

Ps.: Only way to get current position atm I can think of it to get it from ffmpeg output.

[jruggle]

I wanted to add support for aften inside Hybrid but ran into the problem that aften doesn't show progress when I pipe into it from ffmpeg.

Aften needs to know the total number of samples in order to show progress. It can get it 2 ways. For raw audio it is determined from the file size. For wav, aiff, etc, it is determined from the header. When ffmpeg sends output to a pipe, the file size cannot determined. When ffmpeg writes a wav file, the header is not updated with the data size until the end of encoding.

It appears that sox actually writes a valid data size in the header even when streaming to a pipe, if it knows it. Obviously if you're streaming from ffmpeg to sox to aften then sox cannot know the duration either so it cannot pass it to aften.

One workaround might be to add an option to aften so the user can specify the total number of input samples, overriding information from the file size or header if present. Then your program or script could get the duration from ffmpeg (or ffprobe) and multiply by the sample rate. If that option would be useful, I would consider adding it.

[Selur]

So sample rate * time in seconds = frame count ? If so you don't need to add such an option since I already got the duration in seconds and calculating the progress from the output I get when calling with '-v' as parameter isn't a problem. :)

Cu Selur

[Anima123]

Hi jruggle,

Any plan for introducing psychoacoustic model in your encoder? I am the one who have offered you some e-books at the hydrogenaudio forum.

Regards,

[tebasuna51]

So sample rate * time in seconds = frame count ? If so you don't need to add such an option since I already got the duration in seconds and calculating the progress from the output I get when calling with '-v' as parameter isn't a problem. :)

Cu Selur
Exactly:
sample rate * time in seconds = sample count

Frames, in compressed audio, can have many samples, for instance a frame in standard ac3 have 1536 samples.

[Selur]

Exactly:
sample rate * time in seconds = sample count

Frames, in compressed audio, can have many samples, for instance a frame in standard ac3 have 1536 samples.
when using '-v' aften shows each frame's stats assuming these are ac3 frames:
(sample rate * time in seconds)/1536 would provide me the number of frames aften will show for piped input, right?

Cu Selur

[LigH]

Look at the number closely! 1536 = 0600h.

[Selur]

hups, thanks for the correction! (fixed it in the above post)
-> it's working !!

[jruggle]

Any plan for introducing psychoacoustic model in your encoder? I am the one who have offered you some e-books at the hydrogenaudio forum.

Eventually. I got it working once but I wasn't satisfied with it. I have very limited time, but it's still on my todo list.

[Vincent Vega]

can't find an explanation for encoder output log, like these lines i get in megui:

progress: 100% | q: 333.5 | bw: 38.0 | bitrate: 192.0 kbps
progress: 100% | q: 284.0 | bw: 38.0 | bitrate: 192.0 kbps

what does q and bw (bitwriter?) mean? somewhere i can read about it?

[Abradoks]

what does q and bw (bitwriter?) mean? somewhere i can read about it?
Quality and bandwidth. Try "aften -longhelp".

[Vincent Vega]

thanks man

[video_magic]

Just found this at HA.org...and it sounds promizing. :)


Wisodev's builds :: http://win32builds.sourceforge.net/aften/index.html
For whose who prefer GUIs instead of command lines, I've made a small one (http://kurtnoise.free.fr/index.php?dir=Aften/&file=AftenGUI-1.4.zip).


In this first post, I have noticed that the wizodev link is a 404. Could you update to point to:
http://win32builds.sourceforge.net/

That seems current, thanks.

[LigH]

Please note:

http://sourceforge.net/projects/win32builds/files/
As of 2008-05-10 0:00:00 GMT, this project is no longer under active development.

And http://kurtnoise.free.fr is just as dead.

So - where is a really "current" alternative?