Some years ago I conducted a number of telephone interviews for a book
I'm working on. They were recorded via one of those Radio Shack
interface devices into a low cost cassette portable, with ALC.

When the subject talked, it was generally OK. When I talked, the ALC
would kick in fairly quickly and bring my level down. But when the
subject responded, there was quite a delay in release time, so the
first few words of response were quite low, for maybe 10 seconds.

I need to improve those tapes for transcribing. Fidelity is not an
issue; intelligiblity is.

My plan to "reverse" the ALC is to route them through a compressor, set
the threshhold just above the quietest parts, so that they pass
unaffected, while the louder parts are squished down to their level.
Then I'll push the compressor output level way up and re-dub them to
new cassettes.

And to increase intelligibility, I'll also run it through an equalizer,
or possibly an Aphex.

Am I on the right track here? Is there a variety of compressor that
will actually increase the level of the quiet parts while decreasing
the level of the louder parts?

And which will give me better intelligiblilty: EQ or Aphex?

Help!

wrote:

My plan to "reverse" the ALC is to route them through a compressor, set
the threshhold just above the quietest parts, so that they pass
unaffected, while the louder parts are squished down to their level.
Then I'll push the compressor output level way up and re-dub them to
new cassettes.

What you really need to use is an expander rather than a compressor,
but I really don't think you'll get very effect "uncompression"
automatically. Probably the best thing to do is load the audio (as is)
into a DAW, and then work on the unintelligible parts individually,
highlighting a section, boosting the volume, then working with EQ
and/or a noise reduction tool to reduce the hiss which will be present.
There's nothing you can do about the rise in ambient noise, but
hopefully you can pull words out of that.

It's a time-consuming job. That's why forensic audio specialists get
the big bucks. But you probably have most of the tools you need if you
have any of the popular, even freeware, DAW programs. You might as well
keep a word processor open along with the DAW and just transcribe as
you go. That way you can catch the words when they're fresh in your
mind, and can easily go back to check context.

Unless you need to transcribe many hours of recordings, it might be easier
simply to "ride gain" manually.

On 29 May 2006 06:28:34 -0700, "
wrote:

Some years ago I conducted a number of telephone interviews for a book
I'm working on. They were recorded via one of those Radio Shack
interface devices into a low cost cassette portable, with ALC.

When the subject talked, it was generally OK. When I talked, the ALC
would kick in fairly quickly and bring my level down. But when the
subject responded, there was quite a delay in release time, so the
first few words of response were quite low, for maybe 10 seconds.


I have found a useful way to deal fix similar problems using
Audition, but any DAW program could probably do this. I select the
area from the point that it starts out too quiet to the point where it
is finally OK. I then process a fade in but set the initial volume at
+6dB or +9dB or what ever it needs to be at the start and 0 dB at the
end.

I need to improve those tapes for transcribing. Fidelity is not an
issue; intelligiblity is.

My plan to "reverse" the ALC is to route them through a compressor, set
the threshhold just above the quietest parts, so that they pass
unaffected, while the louder parts are squished down to their level.
Then I'll push the compressor output level way up and re-dub them to
new cassettes.

Is a DAW not available?

And which will give me better intelligiblilty: EQ or Aphex?

eq imo.

Julian

" wrote in
oups.com:

Some years ago I conducted a number of telephone interviews for a book
I'm working on. They were recorded via one of those Radio Shack
interface devices into a low cost cassette portable, with ALC.

When the subject talked, it was generally OK. When I talked, the ALC
would kick in fairly quickly and bring my level down. But when the
subject responded, there was quite a delay in release time, so the
first few words of response were quite low, for maybe 10 seconds.

I need to improve those tapes for transcribing. Fidelity is not an
issue; intelligiblity is.

My plan to "reverse" the ALC is to route them through a compressor, set
the threshhold just above the quietest parts, so that they pass
unaffected, while the louder parts are squished down to their level.
Then I'll push the compressor output level way up and re-dub them to
new cassettes.

And to increase intelligibility, I'll also run it through an equalizer,
or possibly an Aphex.

Am I on the right track here? Is there a variety of compressor that
will actually increase the level of the quiet parts while decreasing
the level of the louder parts?

Yes, a correctly set compressor will bring the loud levels down to the
quiet levels. However, the noise will pump with the levels.

If you crank the levels higher manually during the quiet moments can you
understand the subject, or is the noise level too high? If noise is an
issue, then the compressor won't do the whole job. You'll have to apply
some kind of noise reduction first.

And which will give me better intelligiblilty: EQ or Aphex?

Given the response of the average telephone, I don't see that EQ will do
much good. A phone circuit already emphasizes the frequencies that carry
the most speech information. The best you could do would be to lop off
noise above and below the phone frequencies. A cheap cassette already
does that, too.

An Aphex Compellor *IS* a compresser. Or did you have some other Aphex
device in mind?

A few responses:

There are quite a few tapes, and manual gain riding is out of the
question. My only DAW is a Tascam 2488, but it seems like a lot of time
involved in "ingesting", affecting each "bad spot", and dumping out. It
HAS to go to cassette: I already have them on CD, but if I need to back
up and hit the button wrong, it goes to the beginning of the track -
beginning of the conversation! Why oh why don't CD burners allow you to
put a cue point every minute, like DVD burners?!?

Carey Carlan wrote:
Yes, a correctly set compressor will bring the loud levels down to the
quiet levels. However, the noise will pump with the levels.

If you crank the levels higher manually during the quiet moments can you
understand the subject, or is the noise level too high? If noise is an
issue, then the compressor won't do the whole job. You'll have to apply
some kind of noise reduction first.

I don't care about noise. I expect it.


And which will give me better intelligiblilty: EQ or Aphex?

Given the response of the average telephone, I don't see that EQ will do
much good. A phone circuit already emphasizes the frequencies that carry
the most speech information. The best you could do would be to lop off
noise above and below the phone frequencies. A cheap cassette already
does that, too.

My hearing is getting bad, and I NEED high end. The cassette sounds a
bit muffled. I just need to add some sizzle at the top for
intelligibility.

An Aphex Compellor *IS* a compresser. Or did you have some other Aphex
device in mind?

Yeah, I was talking about an old outboard Type C Exciter. I have 3 of
the things. Anybody want one? For compressors I have an old Boss
RCL-10, which I love, and the Behringer 4-banger.

I think my theory is sound: Set the threshhold at the lower sounds, so
that everything louder will be squished down to that level. At least I
can listen to it in my car to determine which spots I need to
transcribe. Now, I can't hear the quieter parts in my car.

" wrote in
oups.com:

Given the response of the average telephone, I don't see that EQ will
do much good. A phone circuit already emphasizes the frequencies
that carry the most speech information. The best you could do would
be to lop off noise above and below the phone frequencies. A cheap
cassette already does that, too.

My hearing is getting bad, and I NEED high end. The cassette sounds a
bit muffled. I just need to add some sizzle at the top for
intelligibility.

There is no high end in a phone circuit. It tops out at 4-5 kHz.
Anything near that is crackly. Above that is just hiss and noise. No
sizzle on a phone. Give the EQ a try, but don't expect much.

An Aphex Compellor *IS* a compresser. Or did you have some other
Aphex device in mind?

Yeah, I was talking about an old outboard Type C Exciter. I have 3 of
the things. Anybody want one? For compressors I have an old Boss
RCL-10, which I love, and the Behringer 4-banger.

An Exciter shouldn't help much, again because of the lack of any real
treble. Again, let your ears decide.

I think my theory is sound: Set the threshhold at the lower sounds, so
that everything louder will be squished down to that level. At least I
can listen to it in my car to determine which spots I need to
transcribe. Now, I can't hear the quieter parts in my car.

And why are you transcribing in the car?

wrote:
There are quite a few tapes, and manual gain riding is out of the
question.

Why? No money in it?

My only DAW is a Tascam 2488, but it seems like a lot of time
involved in "ingesting", affecting each "bad spot", and dumping out.

That isn't the right kind of DAW to use for this job. You need a
computer, and given the poor quality of the input, you needn't worry
about a high grade sound card or A/D converter. There are many
inexpensive or free programs that have the tools needed to do the job.
Your problem, I guess, is time or the unwillingness for a cllient to
pay for the job done properlyt (or spend a lot of your own time without
being paid).

HAS to go to cassette: I already have them on CD, but if I need to back
up and hit the button wrong, it goes to the beginning of the track -
beginning of the conversation! Why oh why don't CD burners allow you to
put a cue point every minute, like DVD burners?!?

You can do this with Nero, and probably most other CD-writing
computer-based tools, but to get there from where you are now, you
first have to "rip" the CD to create a WAV file and then insert markers
wherever you want them. One every time the speaker changes, or when
there's a pause in the speech would be reasonable. It helps to listen
to where you're placing the markers, but you can get them in the
ballpark very quickly by simply looking at the waveform envelope. Given
the characteristic level change when changing from one speaker to the
other, this should be very easy with your recordings.

My hearing is getting bad, and I NEED high end. The cassette sounds a
bit muffled. I just need to add some sizzle at the top for
intelligibility.

I suspect that there simply isn't any high end on the recording given
that it's a cassette, and didn't you say the interview was over the
telephone? Boosting what's not there won't put it there.

The job you need to do isn't trivial and it's not easy to automate and
get good results. But the tools to do it right are available and not
very expensive. The only thing you need is the time for the manual
operations necessary to get the best results.

I'm sure that in all of this time you have tried your "theory" and can
report back that either it was good enough or that it wasn't. What's
the scoop? Are you not even willing or able to try a simple experiment?
That's how you learn what works and what doesn't.

Carey Carlan wrote:

And why are you transcribing in the car?

Maybe he's in one of those states where it's illegal to talk on a cell
phone while driving, and he needs something to keep his mind off the
traffic.

Mike Rivers wrote:
Carey Carlan wrote:

And why are you transcribing in the car?

Maybe he's in one of those states where it's illegal to talk on a cell
phone while driving, and he needs something to keep his mind off the
traffic.

Oh, you wise guys, you!

A long time ago, I dubbed the cassettes staright to CD, so I could
listen to them in my car, for the following reasons:

1) To determine which interview is on which tape. To complile a TOC, as
it were.

2) A great deal of interview material is useless. As I drive and
listen, I can say oh, yes, I need to transcribe that part, and make
note of the approximate time of the part.


BTW, I've been FIRST in my neighborhood on many techie things: Cobra
cordless phone in 1975, Quadraphonic in 1973, 4-track Teac in 1972,
first VCR in 1980, video camera, DAT, computer in 1980? (OK, it was a
Commodore)

But I've somehow resisted having a pager or cell phone! Sometimes I
don't WANT to be found!

Mike Rivers wrote:
wrote:
There are quite a few tapes, and manual gain riding is out of the
question.

Why? No money in it?

No, I have about 20 90-min tapes.


My only DAW is a Tascam 2488, but it seems like a lot of time
involved in "ingesting", affecting each "bad spot", and dumping out.

That isn't the right kind of DAW to use for this job. You need a
computer, and given the poor quality of the input, you needn't worry
about a high grade sound card or A/D converter. There are many
inexpensive or free programs that have the tools needed to do the job.
Your problem, I guess, is time or the unwillingness for a cllient to
pay for the job done properlyt (or spend a lot of your own time without
being paid).

With a young family and a TV station to run, I couldn't find time even
if I was paid!


HAS to go to cassette: I already have them on CD, but if I need to back
up and hit the button wrong, it goes to the beginning of the track -
beginning of the conversation! Why oh why don't CD burners allow you to
put a cue point every minute, like DVD burners?!?

You can do this with Nero, and probably most other CD-writing
computer-based tools, but to get there from where you are now, you
first have to "rip" the CD to create a WAV file and then insert markers
wherever you want them. One every time the speaker changes, or when
there's a pause in the speech would be reasonable. It helps to listen
to where you're placing the markers, but you can get them in the
ballpark very quickly by simply looking at the waveform envelope. Given
the characteristic level change when changing from one speaker to the
other, this should be very easy with your recordings.

You know, I'm cool with the change from mechanical to computers. I have
Apple G5s with Final Cut Pro, and I'm OK with it. it just seems like
drudgery to input all this stuff and tweak it in detail. I know, I'm a
lazy *******.


My hearing is getting bad, and I NEED high end. The cassette sounds a
bit muffled. I just need to add some sizzle at the top for
intelligibility.

I suspect that there simply isn't any high end on the recording given
that it's a cassette, and didn't you say the interview was over the
telephone? Boosting what's not there won't put it there.

I'm not looking for magic, but to my ears, even goosing the treble on
my car CD player makes them easier to understand.

The job you need to do isn't trivial and it's not easy to automate and
get good results. But the tools to do it right are available and not
very expensive. The only thing you need is the time for the manual
operations necessary to get the best results.

I'm sure that in all of this time you have tried your "theory" and can
report back that either it was good enough or that it wasn't. What's
the scoop? Are you not even willing or able to try a simple experiment?
That's how you learn what works and what doesn't.

HERE's a question. Can I set a compressor to do this?: Set the
threshhold so that it BOOSTS the quiet parts while leaving the louder
stuff over the threshhold uaffected?

Please don't tell me I have to put too much effort into this, Mike! I
don't want to spill my lemonade or drop a bon-bon.

wrote:
There are quite a few tapes, and manual gain riding is out of the
question.

No, I have about 20 90-min tapes.

That sounds like a pretty good $4500 project to me. What's wrong with
that? (20 tapes, 3 hours each, $75/hour)

With a young family and a TV station to run, I couldn't find time even
if I was paid!

That's why you need a subcontractor. Pay some enthusiastic kid $2,000
to do the job and take the family to Six Flags with your $2500 profit.

You know, I'm cool with the change from mechanical to computers. I have
Apple G5s with Final Cut Pro, and I'm OK with it. it just seems like
drudgery to input all this stuff and tweak it in detail. I know, I'm a
lazy *******.

You're right about the drudgery. Computers don't really make our work
any easier, they just change the time we work. We can sometimes work
faster, so we spend more time trying to do a better job. Sometimes we
can, sometimes we can't, but it takes a lot of self-discipline to stop
fooling around when it's good enough, or to use a non-computer process
if that gets the results you need quicker than using a computer.

HERE's a question. Can I set a compressor to do this?: Set the
threshhold so that it BOOSTS the quiet parts while leaving the louder
stuff over the threshhold uaffected?

A compressor doesn't boost anything except at the output. You can
probably find a setting that will reduce the level of the loud parts to
that of the quiet parts, and then make the whole thing louder. But if
there's enough difference in level so that you have trouble hearing the
quiet parts, you're bound to have some unpleasant artifacts introduced
in the loud parts by the gain reduction, and that might reverse your
problem.

Please don't tell me I have to put too much effort into this, Mike! I
don't want to spill my lemonade or drop a bon-bon.

It's easier to ignore the spilled lemonade if you're well paid.

Mike Rivers wrote:
wrote:
There are quite a few tapes, and manual gain riding is out of the
question.

No, I have about 20 90-min tapes.

That sounds like a pretty good $4500 project to me. What's wrong with
that? (20 tapes, 3 hours each, $75/hour)

With a young family and a TV station to run, I couldn't find time even
if I was paid!

That's why you need a subcontractor. Pay some enthusiastic kid $2,000
to do the job and take the family to Six Flags with your $2500 profit.

One apparent bit of confusion: I'm not doing this for a client. it's
for myself, my own book.

$4500.00? I just bought a new furnace. Now, I get in a new Musician's
Friend catalog, and chuck it so I don't get tempted. Go to the store
for a new CD? Fat chance. I have to buy "High School Musical" for the
kidlets. What was I thinking???

wrote:
My hearing is getting bad, and I NEED high end. The cassette sounds a
bit muffled. I just need to add some sizzle at the top for
intelligibility.

What sort of cassette deck are you using for playback? Does it sound
better if you listen to only one channel? How far off IS the azimuth
anyway?
--scott


--
"C'est un Nagra. C'est suisse, et tres, tres precis."

On 30 May 2006 10:26:29 -0700, "
wrote:

$4500.00? I just bought a new furnace. Now, I get in a new Musician's
Friend catalog, and chuck it so I don't get tempted. Go to the store
for a new CD? Fat chance. I have to buy "High School Musical" for the
kidlets. What was I thinking???

There are lots of ways to improve your audio but they all cost time
and or money. If you had a newer copy of Sound Forge the "Wave
Hammer" does a good job at getting everything to the same volume with
minimal distortion to the parts being most heavily compressed. But
then you'd have to buy Sound Forge. It's either time or money or
both.

For minimum time and money, get Sound Forge, record your original
cassettes into a computer using any old sound card you have but make
sure the azimuth is tweaked for maximum treble. If you can't get any
more treble this way than the CD's already have, just rip the CD's
instead and save time. Apply the Wave Hammer. You can quickly find a
setting that will squash your dynamics optimally since you are just
splitting the difference between 2 volumes. Save that setting and
just run a batch conversion on all 20 hours at once. Experiment with
eq. If you find it helpful, save the eq setting and also run a batch.
You requested track marks every minute or so. I don't know how to
automate that part but you can put them in manually. Maybe someone
knows how to automate that part too. Then burn new CD's. I don't
think you'll find an easier way that will sound as good as this.

Julian

wrote:
Some years ago I conducted a number of telephone interviews for a book
I'm working on. They were recorded via one of those Radio Shack
interface devices into a low cost cassette portable, with ALC.

When the subject talked, it was generally OK. When I talked, the ALC
would kick in fairly quickly and bring my level down. But when the
subject responded, there was quite a delay in release time, so the
first few words of response were quite low, for maybe 10 seconds.

I need to improve those tapes for transcribing. Fidelity is not an
issue; intelligiblity is.

My plan to "reverse" the ALC is to route them through a compressor,
set the threshhold just above the quietest parts, so that they pass
unaffected, while the louder parts are squished down to their level.
Then I'll push the compressor output level way up and re-dub them to
new cassettes.

And to increase intelligibility, I'll also run it through an
equalizer, or possibly an Aphex.

Am I on the right track here? Is there a variety of compressor that
will actually increase the level of the quiet parts while decreasing
the level of the louder parts?

And which will give me better intelligiblilty: EQ or Aphex?

Help!

First optimise playback - turn Dolby off, tweak azimuth for best HF response
(even works for telephone material).

Then bandpass filter your material - 200Hz to 5KHz - this will stop the then
stages reacting to out-of-speech-band events.

Then use your compressor with a high comression ratio, fast attack so that
you can hear the quiet stuff whilst not having the loud stuff blow your
already damaged) ears off. Try to set the release time of the compressor to
'match' the ALC to keep things even.

Then use Some EQ - boost from about 2KHz by around 6-9dB - this is where
some of the more sybillant telephone speech frequencies are. Sure, you'll
get lots of hiss but you're willing to put up with this for the sake of
inteilligilibity - you're not broadcasting these interviews, just
transcribing them.

Then add the Aphex. This will synthesize new higher frequency content which
is at least sympathetic to the material. Again, target the higher
frequencies - 3-4KHz.

As you've leveled the whole thing off, you've got precious little dynamic
range left to do any noise reduction and that will probably work against
your intelligibility goal anyway so forget it.

You can record to something like an HHB CD-R830 which allows automatic 1
mintue or 5 minute or 30 minute track IDs or it can work on thresholds.

HTH

Rolo

Scott Dorsey wrote:
wrote:
My hearing is getting bad, and I NEED high end. The cassette sounds a
bit muffled. I just need to add some sizzle at the top for
intelligibility.

What sort of cassette deck are you using for playback?

The tapes were recorded on a RadShack SCR-51 handheld type stereo
cassette recorder, using a direct feed from their 43-236 phone
interface, into the ext mic input. that machine is still available. But
I can also play back on a Sony TC-WE475 dual well.

Does it sound
better if you listen to only one channel?

Nope, no phase cancellation or mistracking that I can hear.

How far off IS the azimuth
anyway?

?? From what reference? This is the machine they were recorded on. I
presume playback matches the azimuth at the time of recording.

BTW, I tried setting my old Boss RCL-10 compressor to push the louder
sounds down close to the level of the quieter sounds, and I goosed the
EQ in the upper mids. It is much easier to hear now! At least I can
make out the words, even against road noise in my car.

wrote:
Scott Dorsey wrote:
wrote:
My hearing is getting bad, and I NEED high end. The cassette sounds a
bit muffled. I just need to add some sizzle at the top for
intelligibility.

What sort of cassette deck are you using for playback?

The tapes were recorded on a RadShack SCR-51 handheld type stereo
cassette recorder, using a direct feed from their 43-236 phone
interface, into the ext mic input. that machine is still available. But
I can also play back on a Sony TC-WE475 dual well.

I'd suggest trying the dual well machine and tweaking the azimuth by
hand. It will be a major pain compared with a real transcription deck,
but I bet you'll get a lot more highe dn.

Does it sound
better if you listen to only one channel?

Nope, no phase cancellation or mistracking that I can hear.

You sure about that? I have really never heard a cassette that didn't
have some cancellation, even when played on the original machine.

BTW, I tried setting my old Boss RCL-10 compressor to push the louder
sounds down close to the level of the quieter sounds, and I goosed the
EQ in the upper mids. It is much easier to hear now! At least I can
make out the words, even against road noise in my car.

You may find a severe bass cut will help more than adding a presence peak,
but sometimes you might need to add both.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

Thanks, Rolo. Just what I was looking for!



Rolo Tomassi wrote:
First optimise playback - turn Dolby off, tweak azimuth for best HF response
(even works for telephone material).

Then bandpass filter your material - 200Hz to 5KHz - this will stop the then
stages reacting to out-of-speech-band events.

Then use your compressor with a high comression ratio, fast attack so that
you can hear the quiet stuff whilst not having the loud stuff blow your
already damaged) ears off. Try to set the release time of the compressor to
'match' the ALC to keep things even.

Then use Some EQ - boost from about 2KHz by around 6-9dB - this is where
some of the more sybillant telephone speech frequencies are. Sure, you'll
get lots of hiss but you're willing to put up with this for the sake of
inteilligilibity - you're not broadcasting these interviews, just
transcribing them.

Then add the Aphex. This will synthesize new higher frequency content which
is at least sympathetic to the material. Again, target the higher
frequencies - 3-4KHz.

As you've leveled the whole thing off, you've got precious little dynamic
range left to do any noise reduction and that will probably work against
your intelligibility goal anyway so forget it.

You can record to something like an HHB CD-R830 which allows automatic 1
mintue or 5 minute or 30 minute track IDs or it can work on thresholds.

HTH

Rolo