[Industrial One]

Is there any application that can convert songs with a really low
sample rate (8 khz) to 44.1 by extrapolating/cloning lower bands into
the missing higher frequency shelf?

[David Nebenzahl]

On 7/29/2010 8:29 PM Industrial One spake thus:

Is there any application that can convert songs with a really low
sample rate (8 khz) to 44.1 by extrapolating/cloning lower bands into
the missing higher frequency shelf?

Can't work. (More properly, I should say that it can't yield any better
result than the original 8kHz recording.)

Think digital pictures. Think of a little thumbnail, say 150 x 250
pixels that you resize to 1500 x 2500 pixels. It'll be bigger, but won't
look any better than the little original. Basically the same thing you'd
be doing with the audio file.


--
The fashion in killing has an insouciant, flirty style this spring,
with the flaunting of well-defined muscle, wrapped in flags.

- Comment from an article on Antiwar.com (http://antiwar.com)

[Mr.T]

"David Nebenzahl" wrote in message
.com...
On 7/29/2010 8:29 PM Industrial One spake thus:

Is there any application that can convert songs with a really low
sample rate (8 khz) to 44.1 by extrapolating/cloning lower bands into
the missing higher frequency shelf?

Can't work. (More properly, I should say that it can't yield any better
result than the original 8kHz recording.)

Of course you *can* add extrapolated information with harmonic synthesis,
whether it sounds "better" will be purely subjective of course. It's common
to do the same thing at the other end using sub-harmonic synthesis too.


Think digital pictures. Think of a little thumbnail, say 150 x 250
pixels that you resize to 1500 x 2500 pixels. It'll be bigger, but won't
look any better than the little original. Basically the same thing you'd
be doing with the audio file.

That also fails to realise that proper interpolation WILL make the picture
look far smoother, and edge detail can be enhanced as well while your at it.
All in all the picture WILL look better to most people, even if no REAL
detail has been added. Same thing with audio.

What you can't do is make a silk purse from a sows ear, or do what the TV
forensics shows would have you believe is possible :-)

MrT.

[Industrial One]

On Jul 29, 10:23 pm, David Nebenzahl wrote:
On 7/29/2010 8:29 PM Industrial One spake thus:

Is there any application that can convert songs with a really low
sample rate (8 khz) to 44.1 by extrapolating/cloning lower bands into
the missing higher frequency shelf?

Can't work. (More properly, I should say that it can't yield any better
result than the original 8kHz recording.)

Think digital pictures. Think of a little thumbnail, say 150 x 250
pixels that you resize to 1500 x 2500 pixels. It'll be bigger, but won't
look any better than the little original. Basically the same thing you'd
be doing with the audio file.

--
The fashion in killing has an insouciant, flirty style this spring,
with the flaunting of well-defined muscle, wrapped in flags.

- Comment from an article on Antiwar.com (http://antiwar.com)

I'm aware it's not possible to restore something out of nothing.
That's not what I asked for. I asked if any program exists that tries
to replicate the missing higher frequencies by extrapolating from the
audio that already exists. Pseudo-high quality, pseudorestoration,
call it what the **** you want.

Example:

http://i26.tinypic.com/35atabp.jpg Song at 12 khz
http://i32.tinypic.com/intlw6.jpg Song at 22 khz (upper frequencies
restored)

Notice how the pattern is predictable. Any program that does even a
half-ass job of this is desirable. Anyone got any leads what to look
for?

[Doug McDonald[_4_]]

On 7/30/2010 7:34 AM, Dick Pierce wrote:
Industrial One wrote:

I'm aware it's not possible to restore something out of nothing.

Then that should be the end of it.

That's not what I asked for.

Actually, you did.

I asked if any program exists that tries
to replicate the missing higher frequencies by extrapolating from the
audio that already exists. Pseudo-high quality, pseudorestoration,
call it what the **** you want.



Does not HD radio actually do this: "spectral band replication"?

Doug McDonald

[David Nebenzahl]

On 7/29/2010 10:34 PM Industrial One spake thus:

I'm aware it's not possible to restore something out of nothing.
That's not what I asked for. I asked if any program exists that tries
to replicate the missing higher frequencies by extrapolating from the
audio that already exists. Pseudo-high quality, pseudorestoration,
call it what the **** you want.

Example:

http://i26.tinypic.com/35atabp.jpg Song at 12 khz
http://i32.tinypic.com/intlw6.jpg Song at 22 khz (upper frequencies
restored)

What is that display even showing? Pardon my ignorance, but is that some
kind of spectrum analyzer? Or just a waveform viewer? And what does it
mean that the top quarter of the display is chopped off in the "before"
view but visible in the "after" one? Looks suspicious to me ...


--
The fashion in killing has an insouciant, flirty style this spring,
with the flaunting of well-defined muscle, wrapped in flags.

- Comment from an article on Antiwar.com (http://antiwar.com)

[Don Pearce[_3_]]

On Fri, 30 Jul 2010 11:38:01 -0700, David Nebenzahl
wrote:

On 7/29/2010 10:34 PM Industrial One spake thus:

I'm aware it's not possible to restore something out of nothing.
That's not what I asked for. I asked if any program exists that tries
to replicate the missing higher frequencies by extrapolating from the
audio that already exists. Pseudo-high quality, pseudorestoration,
call it what the **** you want.

Example:

http://i26.tinypic.com/35atabp.jpg Song at 12 khz
http://i32.tinypic.com/intlw6.jpg Song at 22 khz (upper frequencies
restored)

What is that display even showing? Pardon my ignorance, but is that some
kind of spectrum analyzer? Or just a waveform viewer? And what does it
mean that the top quarter of the display is chopped off in the "before"
view but visible in the "after" one? Looks suspicious to me ...

No, the upper, HF portion of the display is visible in the "before"
view. It has been chopped off in the after view. So this is not a
picture of a restoration, but of a high cut.

You can tell which is which by the info box at the bottom which says
"initial state" when the HF is there, and "silence" when it isn't.

I'm afraid Mr Industrial is a purveyor of porky pies.

d

[Industrial One]

On Jul 30, 6:34 am, Dick Pierce wrote:
Industrial One wrote:


I'm aware it's not possible to restore something out of nothing.

Then that should be the end of it.

That's not what I asked for.

Actually, you did.

I asked if any program exists that tries
to replicate the missing higher frequencies by extrapolating from the
audio that already exists. Pseudo-high quality, pseudorestoration,
call it what the **** you want.

Okay wise-ass, there is a difference between "restoring something out
of nothing" and "restoring something out of something." That better?

But whatever, I upsampled via a resampler with no anti-aliasing and
minimum accuracy and the results aren't too bad actually. I didn't
know it was this easy.

It works well on 22 kHz to 44 but sucks with trying to restore ultra-
low quality like the piece of audio I had which was at 6 kHz. I'm sure
there is something more advanced out there?



On Jul 30, 12:38*pm, David Nebenzahl wrote:
What is that display even showing? Pardon my ignorance, but is that some
kind of spectrum analyzer? Or just a waveform viewer? And what does it
mean that the top quarter of the display is chopped off in the "before"
view but visible in the "after" one? Looks suspicious to me ...

A song in spectrographic view. The horizontal blips you see are piano
notes. And the solid vertical square stumps are the drums. What you
see chopped off is the higher frequency shelf, which I deleted to give
you guys a chance to use your imagination and notice how easily
predictable patterns it has, and how high quality frequency
restoration can be feasible.

[Mr.T]

"Dick Pierce" wrote in message
...
Sure. Go down to your local radio shack. Buy a couple of
small signal diodes. Put 'em on the output of your player.

Or, unglue the surrounds of your speakers, misalign them and
reglue them.

Or, write a simple little program that simply resamples at
a higher rate with no filtering at all. It's real simple
to do. Or just take every sample and square it or take
its square root, or some other non-linear function.

All three methods will generate predictable high frequencies
where there were none before. All three are absolutely
garaunteed to do a half-ass job of it. Thus, all three
would meet what seems to be one of your most important
requirements.


For once I disagree with you Dick. Good spectral synthesis will sound a lot
better than simple harmonic distortion. Whether it sounds any better than
the original (B/W limited version) is definitely another matter and purely
*subjective* of course. But let's not forget how popular the Aphex Aural
Exciter once was, and it's a lot easier to do a far better job these days
with digital techniques. (and yes I know it uses phase shifting etc. as well
as spectral synthesis, but the point remains I think that what *some* people
consider an "improvement" can be achieved, if that's all he is after)

MrT.

[Kevin McMurtrie[_3_]]

In article ,
"Mr.T" MrT@home wrote:

"Dick Pierce" wrote in message
...
Sure. Go down to your local radio shack. Buy a couple of
small signal diodes. Put 'em on the output of your player.

Or, unglue the surrounds of your speakers, misalign them and
reglue them.

Or, write a simple little program that simply resamples at
a higher rate with no filtering at all. It's real simple
to do. Or just take every sample and square it or take
its square root, or some other non-linear function.

All three methods will generate predictable high frequencies
where there were none before. All three are absolutely
garaunteed to do a half-ass job of it. Thus, all three
would meet what seems to be one of your most important
requirements.


For once I disagree with you Dick. Good spectral synthesis will sound a lot
better than simple harmonic distortion. Whether it sounds any better than
the original (B/W limited version) is definitely another matter and purely
*subjective* of course. But let's not forget how popular the Aphex Aural
Exciter once was, and it's a lot easier to do a far better job these days
with digital techniques. (and yes I know it uses phase shifting etc. as well
as spectral synthesis, but the point remains I think that what *some* people
consider an "improvement" can be achieved, if that's all he is after)

MrT.

Tonal sound processors are about matching levels between frequency
spectrums in a way that sounds pleasing. They can also inject a
synthesized sound or distortion that tracks a selected component of the
input. One processor is tuned specifically for one source of the music
- one for lead vocals, one or more for backing vocals, one for bass, one
for drums, etc. These processors can't do anything that sounds good on
a final mix.

As far as half-assed solutions go, over-enhancement plus modulated high
frequency random noise does a good job at tricking humans. It works on
photos, videos, voice, and music.

I forgot where I saw it, but there was a software research project that
would attempt to recognize musical components and repair damage using a
library of samples. I suspect it needed tedious amounts of human input
and still produced flaws where the human misjudged art for damage.
--
I won't see Google Groups replies because I must filter them as spam

[Mr.T]

"Kevin McMurtrie" wrote in message
...
Tonal sound processors are about matching levels between frequency
spectrums in a way that sounds pleasing. They can also inject a
synthesized sound or distortion that tracks a selected component of the
input. One processor is tuned specifically for one source of the music
- one for lead vocals, one or more for backing vocals, one for bass, one
for drums, etc. These processors can't do anything that sounds good on
a final mix.

No argument from me, but of course the OP's MP3's don't either, or he
wouldn't be asking how to add high frequencies.


Noise does a good job at tricking humans. It works on
photos, videos, voice, and music.

Most people find it objectionable in photo's, video, voice and music. Only
rarely is it ever added deliberately, and then only by people with weird
ideas about what is "artistic".


I forgot where I saw it, but there was a software research project that
would attempt to recognize musical components and repair damage using a
library of samples. I suspect it needed tedious amounts of human input
and still produced flaws where the human misjudged art for damage.

Yep, no shortage of people who think "art" is anything vastly inferior to
reality.

MrT.

[Doug McDonald[_4_]]

Then let us ask the serious question:

Why would the people who designed the US digital radio system
actually write an "upsampling" system (called "spectral
band replication" into the US standard?

Doug McDonald

[Kevin McMurtrie[_3_]]

In article ,
Doug McDonald wrote:

Then let us ask the serious question:

Why would the people who designed the US digital radio system
actually write an "upsampling" system (called "spectral
band replication" into the US standard?

Doug McDonald

It's part of HE-AAC, which had a lot of research go into hiding extreme
digital losses. It's a smoke and mirrors trick, except for your ears.
If you hear the music for the first time with HE-AAC it sounds passable.
If you hear the lossless version first then HE-AAC, it's obvious that
it's not right.
--
I won't see Google Groups replies because I must filter them as spam

[Mr.T]

"Doug McDonald" wrote in message
...
Why would the people who designed the US digital radio system
actually write an "upsampling" system (called "spectral
band replication" into the US standard?

Because the old adage of "you can fool some of the people all of the time,
and all of the people some of the time" is enough for them.

MrT.

[Arny Krueger]

"Industrial One" wrote in
message


Is there any application that can convert songs with a
really low sample rate (8 khz) to 44.1 by
extrapolating/cloning lower bands into the missing higher
frequency shelf?

I seem to recall that Izotope has some software that synthesizes highs by
leaving some steps out of the usual process of upsampling.

http://www.izotope.com/products/audio/RX/

Just remember, what you want is a syntesized EFX, and not some form of
higher fidelity to the original music.

[Industrial One]

On Aug 2, 6:03*am, "Arny Krueger" wrote:
"Industrial One" wrote in


Is there any application that can convert songs with a
really low sample rate (8 khz) to 44.1 by
extrapolating/cloning lower bands into the missing higher
frequency shelf?

I seem to recall that Izotope has some software that synthesizes highs by
leaving some steps out of the usual process of upsampling.

http://www.izotope.com/products/audio/RX/

Just remember, what you want is a syntesized EFX, and not some form of
higher fidelity to the original music.

I have Izotope, just not the Advanced edition. Know any other?