Re: Reducing noise but keep speech - Soundforge or other?

[S O'Neill]

I've been known to run folks' stuff through SoundSoap on my Mac. How many are
you talking about?

wrote:

> I have some recordings of people speaking, which are obscured by a
> high degree of noise; the recordings being from aircraft, subway
> trains (the Tube), general outdoor/street noises. I need to keep the
> speech intact as much as possible, while reducing noise; said noise
> being generally much louder than the speech.
>
> Is this at all possible? Can an app like Soundforge do it, I see there
> is a Noise reduction option with SF? Are there other specialist
> (Windows) apps which might work, or any audio editing apps which are
> reputable for this application?
>
> thanks......

[gmc]

On Fri, 03 Oct 2003 22:23:35 +0100,
wrote:
>I have some recordings of people speaking, which are obscured by a
>high degree of noise; the recordings being from aircraft, subway
>trains (the Tube), general outdoor/street noises. I need to keep the
>speech intact as much as possible, while reducing noise; said noise
>being generally much louder than the speech.
>
>Is this at all possible? Can an app like Soundforge do it, I see there
>is a Noise reduction option with SF? Are there other specialist
>(Windows) apps which might work, or any audio editing apps which are
>reputable for this application?
>
AFAIK, Soundforge & the simalar Cooledit have functions to sample the
noise & then subtract that from your programme - works best on
continuous, cyclical noise like hum or airconditioning units etc -
helps if u have a clean section of the noise (1-2 sec) so it can
genetate a noise profile from that & then can strip that from the
programme. Expect artifacts like loss of treble & 'underwater voice'
etc if u try to stretch the function too far- think of low bitstream
RealAudio..

You might also try a passband filter - eg brickwall out everything
except from 400Hz up to 2000Hz -get rid of some of the extraneous that
way..


>thanks......

[gmc]

On Fri, 03 Oct 2003 22:23:35 +0100,
wrote:
>I have some recordings of people speaking, which are obscured by a
>high degree of noise; the recordings being from aircraft, subway
>trains (the Tube), general outdoor/street noises. I need to keep the
>speech intact as much as possible, while reducing noise; said noise
>being generally much louder than the speech.
>
>Is this at all possible? Can an app like Soundforge do it, I see there
>is a Noise reduction option with SF? Are there other specialist
>(Windows) apps which might work, or any audio editing apps which are
>reputable for this application?
>
AFAIK, Soundforge & the simalar Cooledit have functions to sample the
noise & then subtract that from your programme - works best on
continuous, cyclical noise like hum or airconditioning units etc -
helps if u have a clean section of the noise (1-2 sec) so it can
genetate a noise profile from that & then can strip that from the
programme. Expect artifacts like loss of treble & 'underwater voice'
etc if u try to stretch the function too far- think of low bitstream
RealAudio..

You might also try a passband filter - eg brickwall out everything
except from 400Hz up to 2000Hz -get rid of some of the extraneous that
way..


>thanks......

[Robert Orban]

Beware of killing intelligibility even if your first impression is that a
given process helps audio quality. Losing low-energy consonants
can definitely make dialog harder to understand.

There seems to be an absolute plague of unintelligible dialog in
current feature film mixing, and one suspect is the various dynamic
noise reduction processes that are popular. I started to doubt my
hearing, except that older films seem to sound just fine.

This trend is a real mystery to me. In the good ol' days, dialog
intelligibility was the #1 goal in motion picture and television
sound.

In article >,
says...
>
>
>I have some recordings of people speaking, which are obscured
by a
>high degree of noise; the recordings being from aircraft, subway
>trains (the Tube), general outdoor/street noises. I need to keep the
>speech intact as much as possible, while reducing noise; said
noise
>being generally much louder than the speech.
>
>Is this at all possible? Can an app like Soundforge do it, I see there
>is a Noise reduction option with SF? Are there other specialist
>(Windows) apps which might work, or any audio editing apps
which are
>reputable for this application?
>
>thanks......

[Robert Orban]

Beware of killing intelligibility even if your first impression is that a
given process helps audio quality. Losing low-energy consonants
can definitely make dialog harder to understand.

There seems to be an absolute plague of unintelligible dialog in
current feature film mixing, and one suspect is the various dynamic
noise reduction processes that are popular. I started to doubt my
hearing, except that older films seem to sound just fine.

This trend is a real mystery to me. In the good ol' days, dialog
intelligibility was the #1 goal in motion picture and television
sound.

In article >,
says...
>
>
>I have some recordings of people speaking, which are obscured
by a
>high degree of noise; the recordings being from aircraft, subway
>trains (the Tube), general outdoor/street noises. I need to keep the
>speech intact as much as possible, while reducing noise; said
noise
>being generally much louder than the speech.
>
>Is this at all possible? Can an app like Soundforge do it, I see there
>is a Noise reduction option with SF? Are there other specialist
>(Windows) apps which might work, or any audio editing apps
which are
>reputable for this application?
>
>thanks......

[Charlie Hubbard]

On Sat, 04 Oct 2003 01:03:58 GMT, (gmc) wrote:
>You might also try a passband filter - eg brickwall out everything
>except from 400Hz up to 2000Hz -get rid of some of the extraneous that
>way..

This could be done relatively quickly in Sound Forge. It never hurts
to try different ideas. However, in my experience, if you set the
high frequency limit this low it will badly distort certain speech
sounds. "S"s suffer in particular. Cutting out the high frequency
components of an "S" make it sound more like a "th"... suddenly
everyone is talking with a lisp! When recording texts for use as
telephone announcements we used to limit the upper frequency to 4000Hz
in preperation to resample to 8kHz (T1 voice data rate is 8,000
samples/second). Depending on who was doing the voice, I would still
occasionally notice lisping.


Charlie Hubbard

[Charlie Hubbard]

On Sat, 04 Oct 2003 01:03:58 GMT, (gmc) wrote:
>You might also try a passband filter - eg brickwall out everything
>except from 400Hz up to 2000Hz -get rid of some of the extraneous that
>way..

This could be done relatively quickly in Sound Forge. It never hurts
to try different ideas. However, in my experience, if you set the
high frequency limit this low it will badly distort certain speech
sounds. "S"s suffer in particular. Cutting out the high frequency
components of an "S" make it sound more like a "th"... suddenly
everyone is talking with a lisp! When recording texts for use as
telephone announcements we used to limit the upper frequency to 4000Hz
in preperation to resample to 8kHz (T1 voice data rate is 8,000
samples/second). Depending on who was doing the voice, I would still
occasionally notice lisping.


Charlie Hubbard

[Charlie Hubbard]

On Fri, 03 Oct 2003 22:23:35 +0100,
wrote:

>I have some recordings of people speaking, which are obscured by a
>high degree of noise; the recordings being from aircraft, subway
>trains (the Tube), general outdoor/street noises. I need to keep the
>speech intact as much as possible, while reducing noise; said noise
>being generally much louder than the speech.
>
>Is this at all possible? Can an app like Soundforge do it, I see there
>is a Noise reduction option with SF? Are there other specialist
>(Windows) apps which might work, or any audio editing apps which are
>reputable for this application?
>
>thanks......

The Sonic Foundry Noise Reduction plug-in for Sound Forge *can* do a
pretty amazing job of eliminating background noise but it depends on
the type of noise. In particular, the technique that SF uses assumes:

1) You have a section of recording (usually in the same piece of
material you're trying to clean up, but not always) that consists of
no other sounds other than the noise you want to remove. For
instance, when recording speech, there is often "silence" before or
after the spoken text that contains nothing but the noise you're
trying to remove.

2) The noise to be removed is constant (at least in its frequency
components and preferably also in its magnitude) across the entire
section of material to be cleaned up. This is often true with things
like air conditioner hiss for instance. It seems like the drone of a
jet engine from inside the cabin is also a good candidate for the type
of noise that can be reduced using the Sound Forge technique. I don't
think it would fare so well against subway noise or street/traffic
noise since it is not constant.

Good luck!


Charlie Hubbard

[Charlie Hubbard]

On Fri, 03 Oct 2003 22:23:35 +0100,
wrote:

>I have some recordings of people speaking, which are obscured by a
>high degree of noise; the recordings being from aircraft, subway
>trains (the Tube), general outdoor/street noises. I need to keep the
>speech intact as much as possible, while reducing noise; said noise
>being generally much louder than the speech.
>
>Is this at all possible? Can an app like Soundforge do it, I see there
>is a Noise reduction option with SF? Are there other specialist
>(Windows) apps which might work, or any audio editing apps which are
>reputable for this application?
>
>thanks......

The Sonic Foundry Noise Reduction plug-in for Sound Forge *can* do a
pretty amazing job of eliminating background noise but it depends on
the type of noise. In particular, the technique that SF uses assumes:

1) You have a section of recording (usually in the same piece of
material you're trying to clean up, but not always) that consists of
no other sounds other than the noise you want to remove. For
instance, when recording speech, there is often "silence" before or
after the spoken text that contains nothing but the noise you're
trying to remove.

2) The noise to be removed is constant (at least in its frequency
components and preferably also in its magnitude) across the entire
section of material to be cleaned up. This is often true with things
like air conditioner hiss for instance. It seems like the drone of a
jet engine from inside the cabin is also a good candidate for the type
of noise that can be reduced using the Sound Forge technique. I don't
think it would fare so well against subway noise or street/traffic
noise since it is not constant.

Good luck!


Charlie Hubbard

[Ricky W. Hunt]

"gmc" > wrote in message
...
> AFAIK, Soundforge & the simalar Cooledit have functions to sample the
> noise & then subtract that from your programme - works best on
> continuous, cyclical noise like hum or airconditioning units etc -

That's true. If you're lucky and can find a good section of voice you can
also approach it from the reverse angle. Sample the voice and set the plugin
to "keep the noise". This can sometimes work in a pinch when you can't get
at it from the other end. From that standpoint the voice may be more
"continuous" than the random noise.

[Ricky W. Hunt]

"gmc" > wrote in message
...
> AFAIK, Soundforge & the simalar Cooledit have functions to sample the
> noise & then subtract that from your programme - works best on
> continuous, cyclical noise like hum or airconditioning units etc -

That's true. If you're lucky and can find a good section of voice you can
also approach it from the reverse angle. Sample the voice and set the plugin
to "keep the noise". This can sometimes work in a pinch when you can't get
at it from the other end. From that standpoint the voice may be more
"continuous" than the random noise.

[Marc Wielage]

On Fri, 3 Oct 2003 18:34:40 -0700, Robert Orban wrote
(in message >):

> There seems to be an absolute plague of unintelligible dialog in
> current feature film mixing, and one suspect is the various dynamic
> noise reduction processes that are popular. I started to doubt my
> hearing, except that older films seem to sound just fine.
>-----------------------------<snip>----------------------------<

There's a trend in modern feature film sound recording today to try to use
the actual production sound dialog as much as possible. I think in the old
days, at least half of the dialog would be looped; in action films (like STAR
WARS in the 1970s), probably 3/4 of the dialog would be looped, maybe more.

But it's a lot different today. Part of it is a creative trend, where the
directors know that the performance from a line looped in an ADR studio will
be markedly different than what was done on location. You can make a good
argument that great actors can actually improve performances this way, but I
suspect that actors are losing this ability as new generations come and go.
And it's hard to argue with the fact that a looped line never sounds quite
the same as one done on location, no matter how much EQ, reverb, or
processing is done to the sound.

And another part is that the equipment has improved to the point where small
microphones, digital wireless transmitters, and digital recorders have made
it much easier to get good dialog quality on location. Multi-track recording
makes it possible to use a plethora of wireless mikes, shotguns, and hidden
mikes, even within the same scene, so that no actor is ever out of range.

But the real problem (IMHO) is speed and lack of post-production time.
Movies are being rushed through so quickly, there often isn't time to bring
an actor in and re-record the dialog. So the poor mixers have to basically
just make do with what they have, even if that means subjecting the location
sound to massive amounts of digital processing to remove generator noise,
clothing rustling, wind noise, etc.

Lastly, there are problems where some inexperienced directors ignore warnings
from experienced mixers and demand excessive levels, sometimes at the expense
of dialog intelligibility. There's no easy answer to this, but well-recorded
movies do still exist, and not all movies sound bad, even today.

--MFW

[Marc Wielage]

On Fri, 3 Oct 2003 18:34:40 -0700, Robert Orban wrote
(in message >):

> There seems to be an absolute plague of unintelligible dialog in
> current feature film mixing, and one suspect is the various dynamic
> noise reduction processes that are popular. I started to doubt my
> hearing, except that older films seem to sound just fine.
>-----------------------------<snip>----------------------------<

There's a trend in modern feature film sound recording today to try to use
the actual production sound dialog as much as possible. I think in the old
days, at least half of the dialog would be looped; in action films (like STAR
WARS in the 1970s), probably 3/4 of the dialog would be looped, maybe more.

But it's a lot different today. Part of it is a creative trend, where the
directors know that the performance from a line looped in an ADR studio will
be markedly different than what was done on location. You can make a good
argument that great actors can actually improve performances this way, but I
suspect that actors are losing this ability as new generations come and go.
And it's hard to argue with the fact that a looped line never sounds quite
the same as one done on location, no matter how much EQ, reverb, or
processing is done to the sound.

And another part is that the equipment has improved to the point where small
microphones, digital wireless transmitters, and digital recorders have made
it much easier to get good dialog quality on location. Multi-track recording
makes it possible to use a plethora of wireless mikes, shotguns, and hidden
mikes, even within the same scene, so that no actor is ever out of range.

But the real problem (IMHO) is speed and lack of post-production time.
Movies are being rushed through so quickly, there often isn't time to bring
an actor in and re-record the dialog. So the poor mixers have to basically
just make do with what they have, even if that means subjecting the location
sound to massive amounts of digital processing to remove generator noise,
clothing rustling, wind noise, etc.

Lastly, there are problems where some inexperienced directors ignore warnings
from experienced mixers and demand excessive levels, sometimes at the expense
of dialog intelligibility. There's no easy answer to this, but well-recorded
movies do still exist, and not all movies sound bad, even today.

--MFW

[Richard Kuschel]

>
>I have some recordings of people speaking, which are obscured by a
>high degree of noise; the recordings being from aircraft, subway
>trains (the Tube), general outdoor/street noises. I need to keep the
>speech intact as much as possible, while reducing noise; said noise
>being generally much louder than the speech.
>
>Is this at all possible? Can an app like Soundforge do it, I see there
>is a Noise reduction option with SF? Are there other specialist
>(Windows) apps which might work, or any audio editing apps which are
>reputable for this application?
>
>thanks......
>
>
>

If the noise is louder than the speech, you are pretty much stuck unless the
noise is outside of the main speech band of 300-3kHz.

Noise reduction systems either work on playing with phase relationslips oe
frequency controlled expansion..

If the noise is louder, it pulls the speech aswell as the noise.
Richard H. Kuschel
"I canna change the law of physics."-----Scotty

[Richard Kuschel]

>
>I have some recordings of people speaking, which are obscured by a
>high degree of noise; the recordings being from aircraft, subway
>trains (the Tube), general outdoor/street noises. I need to keep the
>speech intact as much as possible, while reducing noise; said noise
>being generally much louder than the speech.
>
>Is this at all possible? Can an app like Soundforge do it, I see there
>is a Noise reduction option with SF? Are there other specialist
>(Windows) apps which might work, or any audio editing apps which are
>reputable for this application?
>
>thanks......
>
>
>

If the noise is louder than the speech, you are pretty much stuck unless the
noise is outside of the main speech band of 300-3kHz.

Noise reduction systems either work on playing with phase relationslips oe
frequency controlled expansion..

If the noise is louder, it pulls the speech aswell as the noise.
Richard H. Kuschel
"I canna change the law of physics."-----Scotty

[Bob Olhsson]

In article >, Robert Orban
> wrote:

>This trend is a real mystery to me. In the good ol' days, dialog
>intelligibility was the #1 goal in motion picture and television
>sound.

It's the Pro-Tools "process it because you can" syndrome. Just as in
musical tuning and timing adjustments, the average person using (not to
mention creating) the tools has an oversimplified concept of how
everything works.

--
Bob Olhsson Audio Mastery, Nashville TN 615.385.8051
Mastering, Audio for Picture, Mix Evaluation and Quality Control
http://www.hyperback.com/olhsson.html
Over 40 years making people sound better than they ever imagined!

[Bob Olhsson]

In article >, Robert Orban
> wrote:

>This trend is a real mystery to me. In the good ol' days, dialog
>intelligibility was the #1 goal in motion picture and television
>sound.

It's the Pro-Tools "process it because you can" syndrome. Just as in
musical tuning and timing adjustments, the average person using (not to
mention creating) the tools has an oversimplified concept of how
everything works.

--
Bob Olhsson Audio Mastery, Nashville TN 615.385.8051
Mastering, Audio for Picture, Mix Evaluation and Quality Control
http://www.hyperback.com/olhsson.html
Over 40 years making people sound better than they ever imagined!

[[email protected]]

Sounds like you need something along the lines of forensics. You might give
a look at Diamond Cut software, DCLive or DCart, which has some forensics
applications. --

Joe Salerno
Video Works! Is it working for you?
PO Box 273405 - Houston TX 77277-3405
http://joe.salerno.com
"Richard Kuschel" > wrote in message
...
> >
> >I have some recordings of people speaking, which are obscured by a
> >high degree of noise; the recordings being from aircraft, subway
> >trains (the Tube), general outdoor/street noises. I need to keep the
> >speech intact as much as possible, while reducing noise; said noise
> >being generally much louder than the speech.
> >
> >Is this at all possible? Can an app like Soundforge do it, I see there
> >is a Noise reduction option with SF? Are there other specialist
> >(Windows) apps which might work, or any audio editing apps which are
> >reputable for this application?
> >
> >thanks......
> >
> >
> >
>
> If the noise is louder than the speech, you are pretty much stuck unless
the
> noise is outside of the main speech band of 300-3kHz.
>
> Noise reduction systems either work on playing with phase relationslips oe
> frequency controlled expansion..
>
> If the noise is louder, it pulls the speech aswell as the noise.
> Richard H. Kuschel
> "I canna change the law of physics."-----Scotty

[[email protected]]

Sounds like you need something along the lines of forensics. You might give
a look at Diamond Cut software, DCLive or DCart, which has some forensics
applications. --

Joe Salerno
Video Works! Is it working for you?
PO Box 273405 - Houston TX 77277-3405
http://joe.salerno.com
"Richard Kuschel" > wrote in message
...
> >
> >I have some recordings of people speaking, which are obscured by a
> >high degree of noise; the recordings being from aircraft, subway
> >trains (the Tube), general outdoor/street noises. I need to keep the
> >speech intact as much as possible, while reducing noise; said noise
> >being generally much louder than the speech.
> >
> >Is this at all possible? Can an app like Soundforge do it, I see there
> >is a Noise reduction option with SF? Are there other specialist
> >(Windows) apps which might work, or any audio editing apps which are
> >reputable for this application?
> >
> >thanks......
> >
> >
> >
>
> If the noise is louder than the speech, you are pretty much stuck unless
the
> noise is outside of the main speech band of 300-3kHz.
>
> Noise reduction systems either work on playing with phase relationslips oe
> frequency controlled expansion..
>
> If the noise is louder, it pulls the speech aswell as the noise.
> Richard H. Kuschel
> "I canna change the law of physics."-----Scotty