Nabble - SoX

Re: Request for consulting

2009-12-22T11:39:49Z

Jan Stary wrote:

> On Dec 21 15:53:26, Robert Ferney wrote:
>
>
>> I can't play them back at that quality, mainly because the little itty
>> bitty processor we are using for our embedded system only understands
>> 8 bit, 8khz mono PCM wav files.
>>
>
> Processor does not "understand" any files, it just executes instructions.
> So why cannot you play the originals (16b/44.1k) at the original quality?
> What exactly happens if you do? As in, 'play goodfile.ext'?
>
>

It doesn't sound like the system is a general purpose PC or he would
probably be doing exactly that.

I'd suspect a lightweight processor with firmware on ROM which would
mean the program is immutable, or a custom chip or FPGA in which case
the program is the processor.

--
Daniel Taylor VP Operations Vocal Laboratories, Inc
dtaylor@... 952-941-6580x203

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Sox-users mailing list
Sox-users@...
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Request for consulting

2009-12-22T06:31:55Z

On Dec 21 15:53:26, Robert Ferney wrote:
> The origionals have been professionally recorded in a studio, which is
> why I say from a microphone..

So you do have the original 16b/44.1k files available?

> I can't play them back at that quality, mainly because the little itty
> bitty processor we are using for our embedded system only understands
> 8 bit, 8khz mono PCM wav files.

Processor does not "understand" any files, it just executes instructions.
So why cannot you play the originals (16b/44.1k) at the original quality?
What exactly happens if you do? As in, 'play goodfile.ext'?

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Sox-users mailing list
Sox-users@...
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Possible bug in signal handling

2009-12-21T16:50:45Z

On Mon, Dec 21, 2009 at 12:49 PM, Aaron Grahn <aaron.grahn.999@...> wrote:

When play is running in quiet mode, it does not properly handle a SIGINT.

$ play 1.mp3 2.mp3

Control+C skips to the next track.

But, after

$ play -q 1.mp3 2.mp3

Control+C causes the process to die completely.

// Aaron
Vidi fnordos.

This was intentional design although its open for debate. The thought is that if you've requested to show status then your in front of the display and want to interact. If you've turned off output then your most likely running in script mode and not interactive and a Control+C is really meant to terminate the program.

Chris

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Sox-users mailing list
Sox-users@...
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Request for consulting

2009-12-21T15:42:36Z

# trying to normalize volume and trim extra silence before and after.
sox origional.wav -b 16 -c 1 -s cdquality.wav rate 44100 gain -en -3
contrast silence 1 0.1 -50d pad 0.05 reverse silence 1 0.1 -50d pad
0.05 reverse gain -en -1

This looks reasonable.

# sending it to 8bit 8khz mono pcm wav..
sox cdquality.wav -e unsigned-integer -b 8 -c 1 output.wav rate 8000

This also looks ok. In theory, you might get better performance and results by doing the whole thing (normalize and trim and convert to 8-bit) in one big SoX command, but it probably doesn't matter too much.

The big issue is the dithering applied by SoX. In theory, the dithering should be a good thing. In fact, a really good dither would be the best way to improve things. Unfortunately, it seems that the dithering that is available at 8000 Hz does not do a good job when creating 8-bit output. In a few tests, I found that the audio sounded better with no dither. To remove dither, add the "-D" flag.

sox -D cdquality.wav -e unsigned-integer -b 8 -c 1 output.wav rate 8000

I will also repeat that the biggest thing you could do to improve your audio quality would be to use 16-bit audio. If you can possibly get your software updated to enable 16-bit, u-law, or A-law audio, your results will sound a lot better.

As far as SoX goes, I wonder if a better algorithm exists that would do a better job of dithering to 8-bit. The current algorithm introduces a very noticable hiss. Obviously, 8 bit audio is inherently hard to make sound good, but I'm sure SoX could do a better job. Anybody familiar with this kind of thing? How often do people use SoX to make 8-bit audio output? Is it worth investigation?

Re: Request for consulting

2009-12-21T14:53:26Z

The origionals have been professionally recorded in a studio, which is
why I say from a microphone..

I can't play them back at that quality, mainly because the little itty
bitty processor we are using for our embedded system only understands
8 bit, 8khz mono PCM wav files.

I am using sox to make the translation... ie..

# trying to normalize volume and trim extra silence before and after.
sox origional.wav -b 16 -c 1 -s cdquality.wav rate 44100 gain -en -3
contrast silence 1 0.1 -50d pad 0.05 reverse silence 1 0.1 -50d pad
0.05 reverse gain -en -1

# sending it to 8bit 8khz mono pcm wav..
sox cdquality.wav -e unsigned-integer -b 8 -c 1 output.wav rate 8000

-Robert Ferney ( Kolbe 4357 Demonstrator / Myer Brigs INTJ )

On Sun, Dec 20, 2009 at 10:15 AM, Jan Stary <hans@...> wrote:

> On Dec 20 06:38:53, David Favor wrote:
>> David Graff wrote:
>> > hans@... said:
>> >> Why can't you have them in this quality? How exactly do your 8bit/8kHz
>> >> originate? (Are you playing a CD aloud, and recording it with a
>> >> microphone?)
>> >
>> > On behalf of Robert Ferney: the original digital audio presumably sounds
>> > "good", and the problem would be in how the data are converted from an
>> > original 44k/16-bit pcm sampling to the target system's 8k/8-bit pcm.
>
> Which seems to be by a microphone, which would be the
> first thing I would try to avoid. What exactly
> is the audio path from the good 44k/16b original
> to the bad 8/8 recording?
>
>
> ------------------------------------------------------------------------------
> This SF.Net email is sponsored by the Verizon Developer Community
> Take advantage of Verizon's best-in-class app development support
> A streamlined, 14 day to market process makes app distribution fast and easy
> Join now and get one step closer to millions of Verizon customers
> http://p.sf.net/sfu/verizon-dev2dev
> _______________________________________________
> Sox-users mailing list
> Sox-users@...
> https://lists.sourceforge.net/lists/listinfo/sox-users
>

Possible bug in signal handling

2009-12-21T10:49:12Z

When play is running in quiet mode, it does not properly handle a SIGINT.

$ play 1.mp3 2.mp3

Control+C skips to the next track.

But, after

$ play -q 1.mp3 2.mp3

Control+C causes the process to die completely.

// Aaron
Vidi fnordos.

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Sox-users mailing list
Sox-users@...
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Flac

2009-12-20T18:09:28Z

On Fri, Dec 18, 2009 at 11:32 AM, David Favor <david@...> wrote:

Ron Decline wrote:
> Just d/l's SoX. Trying to swap L/R channels on some flac files, but if I enter "sox file1.flac file2.flac swap" I get an error: sox FAIL formats: no handler for detected file type `flac' (I'm on OS X and I have libFLAC installed).
> --Tony

I'll be building a new OS X version of SoX soon (14.3.1-rc3). I'll check out how flac is working with it.

You may have better luck using http://MacPorts.org as I just submitted
a port patch for both sox 14.3.0 & 14.3.1-rc2 a few days ago, so these
will show up shortly.

Glad to hear this... Any patches that make sense to incorporate into SoX itself?

Once its upgraded, I do suspect MacPorts is better option for most people because of some libltdl issues in 14.3.0 version I released. Also, they appear to MP3 support in their applications.

Chris

Re: SoX man page

2009-12-20T17:35:59Z

On Sat, Dec 19, 2009 at 10:10 AM, Ron Decline <rutlecorps@...> wrote:

I have the PDF version of the manual and the links to the html version, but
is it possible to get a man page version (without compiling)?

You mean you would like it in the unix-ish form for text form?

You can get this unix-ish form directly out of cvs. Find the version for release your using and use its "download" option.

http://sox.cvs.sourceforge.net/viewvc/sox/sox/sox.1?view=log
http://sox.cvs.sourceforge.net/viewvc/sox/sox/soxformat.7?view=log

Re: Request for consulting

2009-12-20T09:15:58Z

On Dec 20 06:38:53, David Favor wrote:
> David Graff wrote:
> > hans@... said:
> >> Why can't you have them in this quality? How exactly do your 8bit/8kHz
> >> originate? (Are you playing a CD aloud, and recording it with a
> >> microphone?)
> >
> > On behalf of Robert Ferney: the original digital audio presumably sounds
> > "good", and the problem would be in how the data are converted from an
> > original 44k/16-bit pcm sampling to the target system's 8k/8-bit pcm.

Which seems to be by a microphone, which would be the
first thing I would try to avoid. What exactly
is the audio path from the good 44k/16b original
to the bad 8/8 recording?

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Sox-users mailing list
Sox-users@...
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Conditional processing

2009-12-20T04:41:07Z

Michael Nemoy wrote:
> Hi
>
> I am using sox as SRC and filter with my Squeezeserver setup.
> However, I would like to keep high-rez files intact..
> High-rez are any file with sample size > 16bit and/or sampling rate > 48KHz
>
> What would you suggest?

Use soxi to extract these values then run your filter for files matching
your criteria.

--
Love feeling your best ever, all day, every day?
Click http://RadicalHealth.com for the easy way!

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Sox-users mailing list
Sox-users@...
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Request for consulting

2009-12-20T04:38:53Z

David Graff wrote:

> hans@... said:
>> Why can't you have them in this quality? How exactly do your 8bit/8kHz
>> originate? (Are you playing a CD aloud, and recording it with a
>> microphone?)
>
> On behalf of Robert Ferney: the original digital audio presumably sounds
> "good", and the problem would be in how the data are converted from an
> original 44k/16-bit pcm sampling to the target system's 8k/8-bit pcm.
>
> Getting anything to sound good in 8k/8-bit pcm is a challenge, but maybe
> some sensible conditioning can be done on the original signal, before
> downsampling and quantizing, that would make the end result less bad...
>
> David Graff

You might post a link to an example file for greatest assistance. It's
really tough to guess at how to start with no example audio to work from.

--
Love feeling your best ever, all day, every day?
Click http://RadicalHealth.com for the easy way!

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Sox-users mailing list
Sox-users@...
https://lists.sourceforge.net/lists/listinfo/sox-users

Conditional processing

2009-12-20T04:31:54Z

Hi

I am using sox as SRC and filter with my Squeezeserver setup.
However, I would like to keep high-rez files intact..
High-rez are any file with sample size > 16bit and/or sampling rate > 48KHz

What would you suggest?

thanks
Michael

Re: SoX man page

2009-12-20T04:03:15Z

I have the PDF version of the manual and the links to the html version, but
is it possible to get a man page version (without compiling)?

The man page is included in the source code download as "sox.1".

Re: Request for consulting

2009-12-20T04:02:15Z

Getting anything to sound good in 8k/8-bit pcm is a challenge, but maybe
some sensible conditioning can be done on the original signal, before
downsampling and quantizing, that would make the end result less bad...

I did a little bit of testing, and it seems that the killer is the 8-bit encoding. At least with SoX, anything created at an 8-bit encoding sounds awful. The quantization noise is pretty bad, and the dithering algorithms available via SoX don't seem to work very well at 8-bit rates.

Re: hd

2009-12-19T16:16:02Z

--- On Sat, 12/19/09, Ron Decline <rutlecorps@...> wrote:

> From: Ron Decline <rutlecorps@...>
> Subject: [SoX-users] hd
> To: sox-users@...
> Date: Saturday, December 19, 2009, 12:34 PM
> What does hd: indicate? (headroom?)
>
> In:0.00% 00:00:00.00 [00:04:44.67] Out:0
> [ |
> ] Clip:0
> In:2.38% 00:00:06.78 [00:04:37.89] Out:299k
> [======|=====-] Clip:0
>
> In:16.7% 00:00:47.65 [00:03:57.02] Out:2.10M
> [======|=====-] Hd:2.7 Clip:0
> In:33.4% 00:01:35.20 [00:03:09.46] Out:4.20M
> [======|=====-] Hd:2.7 Clip:0
> In:50.1% 00:02:22.76 [00:02:21.91] Out:6.30M
> [======|=====-] Hd:1.0 Clip:0
> In:66.9% 00:03:10.31 [00:01:34.36] Out:8.39M
> [======|=====-] Hd:1.0 Clip:0
> In:80.4% 00:03:48.76 [00:00:55.90] Out:10.1M
> [======|======] Hd:1.0 Clip:0
> In:83.7% 00:03:58.24 [00:00:46.43] Out:10.5M
> [======|=====-] Hd:1.0 Clip:0
> In:100% 00:04:44.67 [00:00:00.00] Out:12.6M
> [======|======] Hd:1.0 Clip:0
> Done.
> ------------------------------------------------------------------------------
> This SF.Net email is sponsored by the Verizon Developer
> Community
> Take advantage of Verizon's best-in-class app development
> support
> A streamlined, 14 day to market process makes app
> distribution fast and easy
> Join now and get one step closer to millions of Verizon
> customers
> http://p.sf.net/sfu/verizon-dev2dev

Yes, AFAIR it's headroom.

Regards,
Sergei.

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Sox-users mailing list
Sox-users@...
https://lists.sourceforge.net/lists/listinfo/sox-users

hd

2009-12-19T12:34:56Z

What does hd: indicate? (headroom?)

In:0.00% 00:00:00.00 [00:04:44.67] Out:0 [ | ] Clip:0
In:2.38% 00:00:06.78 [00:04:37.89] Out:299k [======|=====-] Clip:0
In:16.7% 00:00:47.65 [00:03:57.02] Out:2.10M [======|=====-] Hd:2.7 Clip:0
In:33.4% 00:01:35.20 [00:03:09.46] Out:4.20M [======|=====-] Hd:2.7 Clip:0
In:50.1% 00:02:22.76 [00:02:21.91] Out:6.30M [======|=====-] Hd:1.0 Clip:0
In:66.9% 00:03:10.31 [00:01:34.36] Out:8.39M [======|=====-] Hd:1.0 Clip:0
In:80.4% 00:03:48.76 [00:00:55.90] Out:10.1M [======|======] Hd:1.0 Clip:0
In:83.7% 00:03:58.24 [00:00:46.43] Out:10.5M [======|=====-] Hd:1.0 Clip:0
In:100% 00:04:44.67 [00:00:00.00] Out:12.6M [======|======] Hd:1.0 Clip:0
Done.
------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Sox-users mailing list
Sox-users@...
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Request for consulting

2009-12-19T11:18:51Z

> David Graff wrote:

> Getting anything to sound good in 8k/8-bit pcm is a challenge,
> but maybe some sensible conditioning can be done on the
> original signal, before downsampling and quantizing, that
> would make the end result less bad...

Yeah, depending on what "good" means.

8k Hz sample rate will allow for sound intelligibility about
like that of a good 'phone connection. Telephone is limited to
less than 3k Hz.

-- Philip

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Sox-users mailing list
Sox-users@...
https://lists.sourceforge.net/lists/listinfo/sox-users

SoX man page

2009-12-19T08:10:42Z

I have the PDF version of the manual and the links to the html version, but
is it possible to get a man page version (without compiling)?

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Sox-users mailing list
Sox-users@...
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Request for consulting

2009-12-19T08:01:15Z

hans@... said:
> Why can't you have them in this quality? How exactly do your 8bit/8kHz
> originate? (Are you playing a CD aloud, and recording it with a
> microphone?)

On behalf of Robert Ferney: the original digital audio presumably sounds
"good", and the problem would be in how the data are converted from an
original 44k/16-bit pcm sampling to the target system's 8k/8-bit pcm.

Getting anything to sound good in 8k/8-bit pcm is a challenge, but maybe
some sensible conditioning can be done on the original signal, before
downsampling and quantizing, that would make the end result less bad...

David Graff

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Sox-users mailing list
Sox-users@...
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Request for consulting

2009-12-18T15:07:38Z

On Dec 18 15:50:00, Robert Ferney wrote:

> - what sort of signal is this? (music, speech, something else?)
> speech
>
> - by "8 bit", do you mean linear (pcm) 8 bit, or mu-law? (mu-law would
> be better, if your embedded system can use it)
>
> The system is kinda dumb, and only understands PCM, 8 bit, mono 8000 Hz
>
> - what creates the initial digital audio content? (microphone/a2d,
> analog-line-in/a2d, synthesis, other digital source?)
>
> Ultimately a microphone, though by the time I get it, It's typically
> CD quality audio.
>
> - does (can) the original digital audio source have a better fidelity
> than 8khz / 8 bit?
>
> Yes, the originals are PCM, 16 bit, 44100 Hz

Why can't you have them in this quality?
How exactly do your 8bit/8kHz originate?
(Are you playing a CD aloud, and recording it with a microphone?)

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Sox-users mailing list
Sox-users@...
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Request for consulting

2009-12-18T14:50:00Z

David Graff asked some good questions..
Questions:

- what sort of signal is this? (music, speech, something else?)
speech

- by "8 bit", do you mean linear (pcm) 8 bit, or mu-law? (mu-law would
be better, if your embedded system can use it)

The system is kinda dumb, and only understands PCM, 8 bit, mono 8000 Hz

- what creates the initial digital audio content? (microphone/a2d,
analog-line-in/a2d, synthesis, other digital source?)

Ultimately a microphone, though by the time I get it, It's typically
CD quality audio.

- does (can) the original digital audio source have a better fidelity
than 8khz / 8 bit?

Yes, the originals are PCM, 16 bit, 44100 Hz

- can you be a little more specific about what's bad and/or what's
better in terms of perceived signal quality? (maybe a couple sample
files to illustrate "not good" vs. "not bad"?)

The results I'm getting are a little scratchy the "better" results are
somewhat 'tinney' but are more clearly understood.

I'm not quite sure what the proper names for some of the filters are..

the graph / screen-shot that I have depects 0db as a baseline.. and
then a collection of filters..
about three of them describe a bell curve in positive db..
one of them an upside down bell curve in negative db.
one looks like half a bell curve, with a steady db loss after the
inverted peak of the curve..
and then a top and bottom filter that drop exponentially.. eliminating
high and low frequencies.

-Robert Ferney ( Kolbe 4357 Demonstrator / Myer Brigs INTJ )

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Sox-users mailing list
Sox-users@...
https://lists.sourceforge.net/lists/listinfo/sox-users

Request for consulting

2009-12-18T13:16:12Z

I'm using SoX as the back-end sound processor an application, and we
are generating 8 bit 8khz mono wav files with it for use in a very
small embedded system.

The problem is, that the results does not sound all that great..

We had someone try to clean them up, and they used a DIGIRACK EQIII, 1
band and 7 band filters to clean up the sound.

I would like some help duplicating the filter effects using sox, so
that we can automate the process.

The effects looks like they include a low pass, a high pass, and
several band-pass filters, and not being an expert in filters, well...
I'd like some help, so I'm looking for someone willing to do some
consultation..

-Robert Ferney ( Kolbe 4357 Demonstrator / Myer Brigs INTJ )

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Sox-users mailing list
Sox-users@...
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Flac

2009-12-18T09:32:58Z

Ron Decline wrote:
> Just d/l's SoX. Trying to swap L/R channels on some flac files, but if I enter "sox file1.flac file2.flac swap" I get an error: sox FAIL formats: no handler for detected file type `flac' (I'm on OS X and I have libFLAC installed).
> --Tony

You may have better luck using http://MacPorts.org as I just submitted
a port patch for both sox 14.3.0 & 14.3.1-rc2 a few days ago, so these
will show up shortly.

--
Love feeling your best ever, all day, every day?
Click http://RadicalHealth.com for the easy way!

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Sox-users mailing list
Sox-users@...
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Best way to convert wav file into raw vol/freq sample data

2009-12-18T09:31:33Z

Thanks. I'll check these out.

Fmiser wrote:

>> David Favor wrote:
>
>> The more I look at adjusting various spoken word recordings
>> to a common level the more it looks like what I really require
>> to build custom vol + compand effect pipelines for each clip.
>>
>> This is fairly straight forward if I can parse an audio clip
>> into each sample to determine both frequency and volume.
>
> One sample cannot contain any info about frequency
> because since the definition of "frequency" is tied directly to
> "change in amplitude between samples"
>
>> Someone let me know what the best tool is to convert a .wav
>> file into a data stream of volume & frequency of each sample.
>
> So I am, for now, going to ignore frequency and just look at
> level.
>
> PCM audio (which technically is only one type of encoding used
> in a .wav - but is by far the most common) is simply a
> amplitude value for every sample.
>
> You can sort of see this with the following exercise:
>
> sox -n hack.wav synth 1 sin 12000
>
> This will generate a one second long .wav file with 12 kHz
> sine wave. Open hack.wav with a hex editor.
>
> hexedit hack.wav
>
> The pattern of high aplitude followed by low aplitude is, I
> think, obvious.
>
> However, I think you are chasing the wrong rabbit - so to
> speak.
>
> It seems to me your goal is to have a large collection of
> audio files that to human ears _sound_ like they are all at
> the same volume during the speach portions (ignoring the
> applause).
>
> Simple amplitude measurements will not work. Sorry!
>
> Our ears have, in effect, level dependent equalization and
> impulse filtering. This means very short duration sounds can
> have a much higher amplitude and still appear to be the same
> volume as steady sounds.
>
> But that's not all. At different sound pressure levels our
> ears reach differently to low and high frequency - but
> especially low frequency content. The corollary is that low
> frequencies don't contribute as much to our perception of
> volume as mid frequencies.
>
> Other than the applause, what you seem to be wanting to do is
> very similar to what "replay-gain" does.
>
> http://wiki.hydrogenaudio.org/index.php?title=Replaygain
> http://en.wikipedia.org/wiki/Replay_Gain
>
> Do you need the applause to remain in the file? If it were
> removed, it could simplfy the process. *smile* I'm wondering
> if there is an "inverse" to the "silence" effect? One that
> will mute (or drastically reduce the level) of any portion of
> the file that exceedes the threshold for more that a given
> time.
>
> If you find a way to ignore the applause, comparing the RMS
> amplitude and ajusting may work for you. Filtering out
> especially the low frequencies before calculating the RMS
> amplitude may improve the results.
>
> -- Philip
>
> ------------------------------------------------------------------------------
> This SF.Net email is sponsored by the Verizon Developer Community
> Take advantage of Verizon's best-in-class app development support
> A streamlined, 14 day to market process makes app distribution fast and easy
> Join now and get one step closer to millions of Verizon customers
> http://p.sf.net/sfu/verizon-dev2dev
> _______________________________________________
> Sox-users mailing list
> Sox-users@...
> https://lists.sourceforge.net/lists/listinfo/sox-users

--
Love feeling your best ever, all day, every day?
Click http://RadicalHealth.com for the easy way!

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Sox-users mailing list
Sox-users@...
https://lists.sourceforge.net/lists/listinfo/sox-users

Flac

2009-12-17T19:09:11Z

Just d/l's SoX. Trying to swap L/R channels on some flac files, but if I enter "sox file1.flac file2.flac swap" I get an error: sox FAIL formats: no handler for detected file type `flac' (I'm on OS X and I have libFLAC installed).
--Tony

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Sox-users mailing list
Sox-users@...
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Best way to convert wav file into raw vol/freq sample data

2009-12-17T17:26:05Z

> David Favor wrote:

> The more I look at adjusting various spoken word recordings
> to a common level the more it looks like what I really require
> to build custom vol + compand effect pipelines for each clip.
>
> This is fairly straight forward if I can parse an audio clip
> into each sample to determine both frequency and volume.

One sample cannot contain any info about frequency
because since the definition of "frequency" is tied directly to
"change in amplitude between samples"

> Someone let me know what the best tool is to convert a .wav
> file into a data stream of volume & frequency of each sample.

So I am, for now, going to ignore frequency and just look at
level.

PCM audio (which technically is only one type of encoding used
in a .wav - but is by far the most common) is simply a
amplitude value for every sample.

You can sort of see this with the following exercise:

sox -n hack.wav synth 1 sin 12000

This will generate a one second long .wav file with 12 kHz
sine wave. Open hack.wav with a hex editor.

hexedit hack.wav

The pattern of high aplitude followed by low aplitude is, I
think, obvious.

However, I think you are chasing the wrong rabbit - so to
speak.

It seems to me your goal is to have a large collection of
audio files that to human ears _sound_ like they are all at
the same volume during the speach portions (ignoring the
applause).

Simple amplitude measurements will not work. Sorry!

Our ears have, in effect, level dependent equalization and
impulse filtering. This means very short duration sounds can
have a much higher amplitude and still appear to be the same
volume as steady sounds.

But that's not all. At different sound pressure levels our
ears reach differently to low and high frequency - but
especially low frequency content. The corollary is that low
frequencies don't contribute as much to our perception of
volume as mid frequencies.

Other than the applause, what you seem to be wanting to do is
very similar to what "replay-gain" does.

http://wiki.hydrogenaudio.org/index.php?title=Replaygain
http://en.wikipedia.org/wiki/Replay_Gain

Do you need the applause to remain in the file? If it were
removed, it could simplfy the process. *smile* I'm wondering
if there is an "inverse" to the "silence" effect? One that
will mute (or drastically reduce the level) of any portion of
the file that exceedes the threshold for more that a given
time.

If you find a way to ignore the applause, comparing the RMS
amplitude and ajusting may work for you. Filtering out
especially the low frequencies before calculating the RMS
amplitude may improve the results.

-- Philip

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Sox-users mailing list
Sox-users@...
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Best way to convert wav file into raw vol/freq sample data

2009-12-17T10:12:55Z

Jan Stary wrote:

> On Dec 17 10:52:35, David Favor wrote:
>> Jan Stary wrote:
>>> On Dec 17 09:07:25, David Favor wrote:
>>>> The more I look at adjusting various spoken word recordings
>>>> to a common level the more it looks like what I really require
>>>> to build custom vol + compand effect pipelines for each clip.
>>>>
>>>> This is fairly straight forward if I can parse an audio clip
>>>> into each sample to determine both frequency and volume.
>>>>
>>>> Someone let me know what the best tool is to convert a .wav
>>>> file into a data stream of volume & frequency of each sample.
>>> "frequency of each sample" doesn't make any sense.
>> The volume only.
>
> Apparently, you need to read the first few chapters of
> http://www.amazon.com/Principles-Digital-Audio-Ken-Pohlmann/dp/0071441565
> and then look at http://en.wikipedia.org/wiki/File:Pcm.svg for a few minutes.
> (Hint: the sampled value _is_ the amplitude.)

Thanks.

--
Love feeling your best ever, all day, every day?
Click http://RadicalHealth.com for the easy way!

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Sox-users mailing list
Sox-users@...
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Best way to convert wav file into raw vol/freq sample data

2009-12-17T09:08:18Z

On Dec 17 10:52:35, David Favor wrote:

> Jan Stary wrote:
> > On Dec 17 09:07:25, David Favor wrote:
> >> The more I look at adjusting various spoken word recordings
> >> to a common level the more it looks like what I really require
> >> to build custom vol + compand effect pipelines for each clip.
> >>
> >> This is fairly straight forward if I can parse an audio clip
> >> into each sample to determine both frequency and volume.
> >>
> >> Someone let me know what the best tool is to convert a .wav
> >> file into a data stream of volume & frequency of each sample.
> >
> > "frequency of each sample" doesn't make any sense.
>
> The volume only.

Apparently, you need to read the first few chapters of
http://www.amazon.com/Principles-Digital-Audio-Ken-Pohlmann/dp/0071441565
and then look at http://en.wikipedia.org/wiki/File:Pcm.svg for a few minutes.
(Hint: the sampled value _is_ the amplitude.)

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Sox-users mailing list
Sox-users@...
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Best way to convert wav file into raw vol/freq sample data

2009-12-17T08:52:35Z

Jan Stary wrote:

> On Dec 17 09:07:25, David Favor wrote:
>> The more I look at adjusting various spoken word recordings
>> to a common level the more it looks like what I really require
>> to build custom vol + compand effect pipelines for each clip.
>>
>> This is fairly straight forward if I can parse an audio clip
>> into each sample to determine both frequency and volume.
>>
>> Someone let me know what the best tool is to convert a .wav
>> file into a data stream of volume & frequency of each sample.
>
> "frequency of each sample" doesn't make any sense.

The volume only.

--
Love feeling your best ever, all day, every day?
Click http://RadicalHealth.com for the easy way!

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Sox-users mailing list
Sox-users@...
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Best way to convert wav file into raw vol/freq sample data

2009-12-17T07:34:23Z

On Dec 17 09:07:25, David Favor wrote:
> The more I look at adjusting various spoken word recordings
> to a common level the more it looks like what I really require
> to build custom vol + compand effect pipelines for each clip.
>
> This is fairly straight forward if I can parse an audio clip
> into each sample to determine both frequency and volume.
>
> Someone let me know what the best tool is to convert a .wav
> file into a data stream of volume & frequency of each sample.

"frequency of each sample" doesn't make any sense.

> The stream is best text based so it can be passed to another
> program to use the data to prepare sox pipelines.
>
> Seems like all this code is available in sox to implement
> the spectrogram effect and I see no way to expose the raw
> data. If there is then this will most likely work.
>
> Thanks.
>
> --
> Love feeling your best ever, all day, every day?
> Click http://RadicalHealth.com for the easy way!
>
> ------------------------------------------------------------------------------
> This SF.Net email is sponsored by the Verizon Developer Community
> Take advantage of Verizon's best-in-class app development support
> A streamlined, 14 day to market process makes app distribution fast and easy
> Join now and get one step closer to millions of Verizon customers
> http://p.sf.net/sfu/verizon-dev2dev
> _______________________________________________
> Sox-users mailing list
> Sox-users@...
> https://lists.sourceforge.net/lists/listinfo/sox-users

Best way to convert wav file into raw vol/freq sample data

2009-12-17T07:07:25Z

The more I look at adjusting various spoken word recordings
to a common level the more it looks like what I really require
to build custom vol + compand effect pipelines for each clip.

This is fairly straight forward if I can parse an audio clip
into each sample to determine both frequency and volume.

Someone let me know what the best tool is to convert a .wav
file into a data stream of volume & frequency of each sample.
The stream is best text based so it can be passed to another
program to use the data to prepare sox pipelines.

Seems like all this code is available in sox to implement
the spectrogram effect and I see no way to expose the raw
data. If there is then this will most likely work.

Thanks.

--
Love feeling your best ever, all day, every day?
Click http://RadicalHealth.com for the easy way!

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Sox-users mailing list
Sox-users@...
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Guru requested for sox-14.3.1-rc2 + normalize question

2009-12-16T14:44:49Z

Jan Stary wrote:

> On Dec 16 12:53:54, David Favor wrote:
>> Looks like this example out of the man page works well across
>> a variety of audio clips:
>>
>> sox infile outfile compand 0.3,1 6:-70,-60,-20 -5 -90 0.2
>>
>> This is the example of compand for listening to audios in a car.
>
> Yes, but you have thousands of them; the fact that it does what you
> want on the five or ten you have listened to now doesn't guarantee
> nothing. But I have no better idea than to use the compander
> (trying a lot of different compander parameters).
>
> Perhaps you might want to first find out how much the recordings differ.
> Are their dynamic ranges very different? (The more so, the bigger
> problem you have). Do their dynamic ranges become similar once
> you throw the rare loud passages out?

Which is why I asked gurus for the best way to accomplish this.

--
Love feeling your best ever, all day, every day?
Click http://RadicalHealth.com for the easy way!

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Sox-users mailing list
Sox-users@...
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Guru requested for sox-14.3.1-rc2 + normalize question

2009-12-16T12:12:10Z

> David Favor wrote:

> Looks like this example out of the man page works well across
> a variety of audio clips:
>
> sox infile outfile compand 0.3,1 6:-70,-60,-20 -5 -90 0.2
>
> This is the example of compand for listening to audios in a
> car.
>
> I do notice the compand function seems to make the first few
> seconds of all my clips very soft, them seems to 'catch up' and
> compand the rest of the clip to a good listening volume.
>
> Further suggestions are great.

The second to the last value (-90) is the "I'll give you a head
start" value. Try changing it to -60. Or maybe -35. That will
get the compand function started at a value closer to what it
continues with. Experiment to find what works for your files.

And normalizing after running it through compand may also help.

-- Philip

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Sox-users mailing list
Sox-users@...
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Guru requested for sox-14.3.1-rc2 + normalize question

2009-12-16T12:07:05Z

> David Favor wrote:

> The processing I'm looking for is something like:
>
> 1) find the loudest 1% of clip and throw these values out
>
> 2) average the other 99% of the clip
>
> 3) normalize this remaining 99% to a common value so all
> past + future clips can be combined together and played
> at a consistent volume

That sound a lot like what "replay gain" does.

But I'm guessing you don't want the clapping to clip - and so
the "compand" function could be part of your solution.

-- Philip

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Sox-users mailing list
Sox-users@...
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Guru requested for sox-14.3.1-rc2 + normalize question

2009-12-16T11:13:14Z

On Dec 16 12:53:54, David Favor wrote:
> Looks like this example out of the man page works well across
> a variety of audio clips:
>
> sox infile outfile compand 0.3,1 6:-70,-60,-20 -5 -90 0.2
>
> This is the example of compand for listening to audios in a car.

Yes, but you have thousands of them; the fact that it does what you
want on the five or ten you have listened to now doesn't guarantee
nothing. But I have no better idea than to use the compander
(trying a lot of different compander parameters).

Perhaps you might want to first find out how much the recordings differ.
Are their dynamic ranges very different? (The more so, the bigger
problem you have). Do their dynamic ranges become similar once
you throw the rare loud passages out?

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Sox-users mailing list
Sox-users@...
https://lists.sourceforge.net/lists/listinfo/sox-users