[William Sommerwerck]

Got a question...

Once upon a time, movie soundtracks were recorded on 35mm film. Let's say
Fred & Ginger's big number is ready to be performed, and the film has been
cued up. How is the sound film synched with the movie film? If it isn't
synched, then lining up the visual and audible recordings could result in a
error of as much as 1/48 of a second. Not likely audible, but why not have
it perfect?

Am I missing something?

[William Sommerwerck]

Begets another question: flutter ... how did (do) they avoid it?

You don't avoid it. If you're recording a continuous (as opposed to sampled)
signal on a moving medium, you'll have flutter.

Of course, in an FM system, the flutter is converted to sideband noise.

[Scott Dorsey]

William Sommerwerck wrote:

Once upon a time, movie soundtracks were recorded on 35mm film. Let's say
Fred & Ginger's big number is ready to be performed, and the film has been
cued up. How is the sound film synched with the movie film? If it isn't
synched, then lining up the visual and audible recordings could result in a
error of as much as 1/48 of a second. Not likely audible, but why not have
it perfect?

Lots of different ways including a "beep box" with a flashing light that
comes on when it detects the 1KC tone on the playback audio.

But the way Fred and Ginger did it was in the pre-pilot tone days. The
camera and the dubber were locked together with three-phase synchros,
so they both came up to speed together.

For live sound scenes, it was easy. The man in the dubber room would
lock to power, the camera operator would lock to power, the PA would come
out and clap the sticks so there was a reference point for the editor,
and everything would go on.

For miming to playback, it was a little different. The dubber and the
camera would be locked together, and the dubber positioned on a reference
point on the film (either a 2-pop or an X mark or something else the
editor could use as reference). The camera operator would open up the
gate in the camera, mark the film (or punch a V in the side of it with
a nifty little gadget if he had a Mitchell camera), and close the gate
up. Then when the dubber and camera rolled together, they'd both have
a reference point to start out from.

Yes, everything can be off by as much as half a frame, but it can't really
be any closer, really.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Scott Dorsey]

Peter Larsen wrote:
Audio on 35 mm mag film with sprocket holes on same spindle as the movie.
Move audio mag film until sync is correct at the clap.

Begets another question: flutter ... how did (do) they avoid it?

With a massy flywheel... film comes off the lower sprocket, goes onto a
tensioned loop around a flywheel, head or optical assembly is on the
flywheel. 1920s patent from Tri-Ergon in Germany... without it, film
sound would be more or less unworkable.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[William Sommerwerck]

"Scott Dorsey" wrote in message
...
William Sommerwerck wrote:

Once upon a time, movie soundtracks were recorded on 35mm film. Let's say
Fred & Ginger's big number is ready to be performed, and the film has
been
cued up. How is the sound film synched with the movie film? If it isn't
synched, then lining up the visual and audible recordings could result in
a
error of as much as 1/48 of a second. Not likely audible, but why not
have
it perfect?

Lots of different ways including a "beep box" with a flashing light that
comes on when it detects the 1KC tone on the playback audio.

But the way Fred and Ginger did it was in the pre-pilot tone days. The
camera and the dubber were locked together with three-phase synchros,
so they both came up to speed together.

For live sound scenes, it was easy. The man in the dubber room would
lock to power, the camera operator would lock to power, the PA would come
out and clap the sticks so there was a reference point for the editor,
and everything would go on.

For miming to playback, it was a little different. The dubber and the
camera would be locked together, and the dubber positioned on a reference
point on the film (either a 2-pop or an X mark or something else the
editor could use as reference). The camera operator would open up the
gate in the camera, mark the film (or punch a V in the side of it with
a nifty little gadget if he had a Mitchell camera), and close the gate
up. Then when the dubber and camera rolled together, they'd both have
a reference point to start out from.

Yes, everything can be off by as much as half a frame, but it can't really
be any closer, really.

Scott, your answer seemed to make perfect sense -- until the last sentence.
ASSUMING one lined up the sprockets of the sound and image -- which was the
point I was making -- you can't be more than 1/48 of a second off. And if
that slight error doesn't matter, there's no point in synching the camera
and the playback.

Please note that I'm not talking about SPEED, but the starting points.
Obviously, if the speeds don't match, then the image and the sound can
slowly drift apart. But that wasn't what I was asking about.

I do get the point about "declutching" and manually synching the sound with
image during editing.

[Scott Dorsey]

William Sommerwerck wrote:

Scott, your answer seemed to make perfect sense -- until the last sentence.
ASSUMING one lined up the sprockets of the sound and image -- which was the
point I was making -- you can't be more than 1/48 of a second off. And if
that slight error doesn't matter, there's no point in synching the camera
and the playback.

The image is quantized in 1/24 second spots. The shutter is open for
some time, perhaps 1/60 of a second, perhaps as much as 1/30th of a
second if your camera allows that.

Do you want to sychronize a tone to when the shutter opens? When the
shutter closes? Half-way in-between? Because what you're putting into
the image is _everything_ that happens across the whole time it's open.
--scott

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

[William Sommerwerck]

Scott, your answer seemed to make perfect sense -- until the last
sentence.
ASSUMING one lined up the sprockets of the sound and image -- which was
the
point I was making -- you can't be more than 1/48 of a second off. And if
that slight error doesn't matter, there's no point in synching the camera
and the playback.

The image is quantized in 1/24 second spots. The shutter is open for
some time, perhaps 1/60 of a second, perhaps as much as 1/30th of a
second if your camera allows that.

Do you want to sychronize a tone to when the shutter opens? When the
shutter closes? Half-way in-between? Because what you're putting into
the image is _everything_ that happens across the whole time it's open.

Why is it that straightforward questions ALWAYS result in complex,
often-pointless discussions?

If the sound is on sprocketed film or tape, and the camera and the player
are started at the same instant, then the film and the sound recording
should exactly line up, and can be edited without any consideration for
synching them -- other than matching up the sprocket holes at the start.
That was what I was asking about -- is that the way it was done? It isn't
these days, but I still don't have an answer as to how it was done 70 years
ago.

Let's just drop the question. It's not one of great importance, and as
everyone is going to perversely misinterpret the question according to his
own lights, there is no point in going further.

[Jason[_14_]]

In article , says...

William Sommerwerck wrote:

Once upon a time, movie soundtracks were recorded on 35mm film. Let's say
Fred & Ginger's big number is ready to be performed, and the film has been
cued up. How is the sound film synched with the movie film? If it isn't
synched, then lining up the visual and audible recordings could result in a
error of as much as 1/48 of a second. Not likely audible, but why not have
it perfect?

Lots of different ways including a "beep box" with a flashing light that
comes on when it detects the 1KC tone on the playback audio.

But the way Fred and Ginger did it was in the pre-pilot tone days. The
camera and the dubber were locked together with three-phase synchros,
so they both came up to speed together.

For live sound scenes, it was easy. The man in the dubber room would
lock to power, the camera operator would lock to power, the PA would come
out and clap the sticks so there was a reference point for the editor,
and everything would go on.

For miming to playback, it was a little different. The dubber and the
camera would be locked together, and the dubber positioned on a reference
point on the film (either a 2-pop or an X mark or something else the
editor could use as reference). The camera operator would open up the
gate in the camera, mark the film (or punch a V in the side of it with
a nifty little gadget if he had a Mitchell camera), and close the gate
up. Then when the dubber and camera rolled together, they'd both have
a reference point to start out from.

Yes, everything can be off by as much as half a frame, but it can't really
be any closer, really.
--scott

That's fascinating, Scott. Thanks.

[[email protected]]

Thanks, Scott. Your tidbits are often fascinating.

[Mxsmanic]

William Sommerwerck writes:

Why is it that straightforward questions ALWAYS result in complex,
often-pointless discussions?

Because simple question/answer exchanges are easily answered by Google, and
are not very interesting to discuss. Questions that elicit lively discussions
tend to be those with complex and often highly debatable answers.

[William Sommerwerck]

I would be dishonest if I didn't say I learned a thing or two here. But the
original question hasn't been answered.

The question was provoked when I watched "Never Gonna Dance" in "Swing
Time". This long sequence is remarkable for a reason besides the terrific
dancing -- it's shot in just two takes. The second take begins once Fred has
danced Ginger up the stairs.

The music is (presumably) a single strip of optical film. So how is the
second take properly synched so that there is no visual discontinuity? Even
if no attempt is made to match sprocket holes, the film of F&G can't be more
than a half-frame off from the sound. I guess the most-basic question is...
IS THIS VISIBLE?

I don't know. But if I were Fred, I wouldn't want to take chances.

[William Sommerwerck]

"Arny Krueger" wrote in message
...

In practice, 1 frame (1/24 or 1/30 of a second) is *not* subjectively
perfect lip synch. It is in some sense good enough, but just doesn't
look quite right. IME we need 10ms accuracy or better for subjectively
perfect synch.

This is another example of how we are able to easily discern variations
in the timing of events in sampled data that vary by amounts that are far
smaller than the sample rate.

Many surround controllers (and possibly some players) include variable delay
to match up the sound and image. I've never adjusted it, because I've never
seen an obvious mismatch.

I have, however, inadvertently had both the main system (playing the Blu-ray
surround tracks) and the monitor (playing the undecoded two-channel feed) at
the same time. You can hear a slight "echo" effect, because they're not
perfectly synchronized. But neither is subjectively unsynchronized with the
image. There is no apparent "phasiness", so the discrepancy is probably
greater than 20ms (the nominal fusion interval).

[Arny Krueger[_4_]]

"Scott Dorsey" wrote in message
...
William Sommerwerck wrote:

Scott, your answer seemed to make perfect sense -- until the last
sentence.
ASSUMING one lined up the sprockets of the sound and image -- which was
the
point I was making -- you can't be more than 1/48 of a second off. And if
that slight error doesn't matter, there's no point in synching the camera
and the playback.

The image is quantized in 1/24 second spots. The shutter is open for
some time, perhaps 1/60 of a second, perhaps as much as 1/30th of a
second if your camera allows that.

Do you want to sychronize a tone to when the shutter opens? When the
shutter closes? Half-way in-between? Because what you're putting into
the image is _everything_ that happens across the whole time it's open.


In practice, 1 frame (1/24 or 1/30 of a second) is *not* subjectively
perfect lip synch. It is in some sense good enough, but just doesn't look
quite right. IME we need 10 mSec accuracy or better for subjectively
perfect synch.

This is another example of how we are able to easily discern time variations
in the timing of events in sampled data that vary by amounts that are far
smaller than the sample rate.

[Arny Krueger[_4_]]

"William Sommerwerck" wrote in message
...
I would be dishonest if I didn't say I learned a thing or two here. But the
original question hasn't been answered.

The question was provoked when I watched "Never Gonna Dance" in "Swing
Time". This long sequence is remarkable for a reason besides the terrific
dancing -- it's shot in just two takes. The second take begins once Fred
has
danced Ginger up the stairs.

The music is (presumably) a single strip of optical film. So how is the
second take properly synched so that there is no visual discontinuity?
Even
if no attempt is made to match sprocket holes, the film of F&G can't be
more
than a half-frame off from the sound. I guess the most-basic question
is...
IS THIS VISIBLE?

I don't know. But if I were Fred, I wouldn't want to take chances.

Most of the common things we do with video have a film analog, albeit one
that takes a lot more work to accomplish.

If I have two takes of the same scene that I want to merge seamlessly, one
approach is

(1) Put the audio and the first take on the timeline.

(2) Cut and remove just the video at a point that is most closely duplicated
by the second take, and before the parts of the first take I want to get rid
of.

(3) Cut the second take's video at the corresponding point.

(4) Cross-fade or simply butt together the two video tracks.

If I was directing this, I would arrange the action so that there was a
point where editing would be easiest. IOW I would attempt to make it easy
and acted out, that the actors hit the same marks at the same time in each
take.

If you are shooting a scene with more than one camera, this all gets easier
because you are already cutting between the cameras, and so you make one of
those cut points take you to the second take.

[Arny Krueger[_4_]]

"William Sommerwerck" wrote in message
...
"Arny Krueger" wrote in message
...

In practice, 1 frame (1/24 or 1/30 of a second) is *not* subjectively
perfect lip synch. It is in some sense good enough, but just doesn't
look quite right. IME we need 10ms accuracy or better for subjectively
perfect synch.

This is another example of how we are able to easily discern variations
in the timing of events in sampled data that vary by amounts that are far
smaller than the sample rate.

Many surround controllers (and possibly some players) include variable
delay
to match up the sound and image. I've never adjusted it, because I've
never
seen an obvious mismatch.

I have had mismatches and used that feature to correct things.

I have, however, inadvertently had both the main system (playing the
Blu-ray
surround tracks) and the monitor (playing the undecoded two-channel feed)
at
the same time. You can hear a slight "echo" effect, because they're not
perfectly synchronized. But neither is subjectively unsynchronized with
the
image. There is no apparent "phasiness", so the discrepancy is probably
greater than 20ms (the nominal fusion interval).

Yes, and similar things happen when you try to time-synch the alternatives
being evaluated during a well-run listening test involving media players.

[Scott Dorsey]

In article , wrote:
Thanks, Scott. Your tidbits are often fascinating.

Thank you. It's part of getting old, I think. I highly recommend getting
old.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Arkansan Raider]

Scott Dorsey wrote:
In article , wrote:
Thanks, Scott. Your tidbits are often fascinating.

Thank you. It's part of getting old, I think. I highly recommend getting
old.
--scott


But whatever you do, don't grow up.

Life is too important to be taken seriously.


---Jeff

[swanny]

On 24/01/2012 2:56 AM, Scott Dorsey wrote:
In article , wrote:
Thanks, Scott. Your tidbits are often fascinating.

Thank you. It's part of getting old, I think. I highly recommend getting
old.
--scott

As Billy Connolly once said:
"I can't wait to get old and pretend I'm deaf."

[Ralph Barone[_2_]]

Scott Dorsey wrote:
In article , wrote:
Thanks, Scott. Your tidbits are often fascinating.

Thank you. It's part of getting old, I think. I highly recommend getting
old.
--scott

Personally, I've been trying to avoid it, but I couldn't find a better
alternative.

[Marc Wielage[_2_]]

On Sun, 22 Jan 2012 12:32:20 -0800, William Sommerwerck wrote
(in article ):

Scott, your answer seemed to make perfect sense -- until the last sentence.
ASSUMING one lined up the sprockets of the sound and image -- which was the
point I was making -- you can't be more than 1/48 of a second off.
------------------------------snip------------------------------

Actually, every perf is 1/96th of a second, so you can slip that far. Two
perfs is 1/48th of a second. Four perfs would be 1/24th of a second. I have
worked with editors who could literally see 1 perf out. I can tell when
something is a *frame* out of sync, but that's my limit.

In real life, when performing to playback, there's a phenomena I've heard
referred to as "artist's error," where they just mouth the lyrics a little
off, or they make a dance move on the wrong beat. Crap happens. We fix it
in post (or at least try to).

Really, really complicated playback situations, like DREAMGIRLS, required
complex setups because there was sometimes dialog being spoken in the middle
of music playback. I have seen *double* timecode slates -- one row of
numbers for time of day, one for music timecode playback -- used for
situations like this, with a full-time Pro Tools playback operator.

For modern music shows (like GLEE and SMASH), they typically send wireless
timecode music to the slate, the camera shoots that, and the editor types in
those numbers to sync it up to the playback track. That way, they can start
right in the middle of the song and know exactly where they are. With
digital cameras, they can also record a scratch track of the music, which
helps make syncing easier.



If the sound is on sprocketed film or tape, and the camera and the player
are started at the same instant, then the film and the sound recording
should exactly line up, and can be edited without any consideration for
synching them -- other than matching up the sprocket holes at the start.
That was what I was asking about -- is that the way it was done?
------------------------------snip------------------------------

Editors jig and jog all the time. You'd be amazed the number of times they
grab shots that are "technically" not in sync with the music, but they
massage the shot in a way to make it fit the flow of the cuts. This is
particularly applicable to slow-motion and other camera tricks.

I can tell you that, prior to the widespread adoption of timecode for film
sound production, they would often play back on the set from one machine, and
then record new sound for every scene, and then we'd use that reference audio
for syncing later (usually on 35mm mag). That would be the easiest way to do
it back in the 1930s and 1940s, though using optical sound during those days
(until mag hit around 1946-1947).

--MFW

[Don Pearce[_3_]]

On 23 Jan 2012 10:56:33 -0500, (Scott Dorsey) wrote:

In article , wrote:
Thanks, Scott. Your tidbits are often fascinating.

Thank you. It's part of getting old, I think. I highly recommend getting
old.

So much better than the alternative.

d

[Mxsmanic]

Seeing this discussion of synchronization reminds me of something I read in an
excellent book about animation. It seems that Disney animators worked very
hard to animate the mouths of cartoon characters to match what human actors
would do when speaking the same words. They were dismayed to discover that
audiences didn't even seem to ca when they dubbed a cartoon in a different
language, for example, nobody noticed the difference between mouth movements
and speech. Perhaps people don't pay attention to mouth movements if the
speaker isn't a human being in live action (or maybe they just don't look at
all, as long as the mouth is opening and closing).

[Marc Wielage[_2_]]

On Wed, 25 Jan 2012 02:46:53 -0800, Mxsmanic wrote
(in article ):

Perhaps people don't pay attention to mouth movements if the
speaker isn't a human being in live action (or maybe they just don't look
at all, as long as the mouth is opening and closing).
------------------------------snip------------------------------

The general rule of thumb (or mouth) is, as long as the lips stop moving when
the words stop, audiences can buy it... to a point.

Some "lip flap" is expected in animation, but they actually work on that
pretty hard, at least in features.

--MFW

[Arny Krueger[_4_]]

"Mxsmanic" wrote in message
...

Seeing this discussion of synchronization reminds me of something I read
in an
excellent book about animation. It seems that Disney animators worked very
hard to animate the mouths of cartoon characters to match what human
actors
would do when speaking the same words. They were dismayed to discover that
audiences didn't even seem to ca when they dubbed a cartoon in a
different
language, for example, nobody noticed the difference between mouth
movements
and speech. Perhaps people don't pay attention to mouth movements if the
speaker isn't a human being in live action (or maybe they just don't look
at
all, as long as the mouth is opening and closing).

When you say that "nobody noticed" you're probably really saying that nobody
complained. For one reason or the the other I've turned out a fair amount of
stuff that had poor or even really bad lip synch. Most people are still
grateful enough to have sound and motion pictures, whether they are in
perfect synch or not.