Way back in the old days of real motion picture film, sound sync was handled by a tone generator in the camera that would send an audio pulse for each frame shot. This was supplanted by crystal sync devices later on, but many old cameras still use pulse sync.
To use it you’d use one track on a multi-track recorder to record the pulse stream from the camera. On the other track (if 2 track) you’d record dialog. The multitrack recorder would then play back in sync with the pulse stream for the mix down.
So my question is, could Ardour be used to take a pulse stream from a camera recorded to one channel and manipulate playback in sync for edit in an NLE? Film is typically processed and scanned these days. So it would playback at 24 /.23.997 fps in sync. All that’s necessary is to stretch or compress the audio stream a few milliseconds per the reference sync pulse.
This HAS to be possible. (though not necessarily worthwhile lol)
That system is newer, probably from the 80s. The systems I’m talking about are from the 50s - 70s, and they’re dumb. They just send a reference pulse for every frame of some single frequency tone. No meta data or digital encoding involved. The pulse stream would go to a portable reel to reel or 4 track cassette modified for sync pulse playback. And with these old camera motors, there’s considerable drift. No way to sync audio without it.
You still see them in rental houses with old gear, renting out ancient Arri, Bolex, Beauliu, and Braun 16mm / 8mm cameras which were once used for news or short film. They’re generally the cheapest thing to rent if you happen to want a 16mm or S8mm sync sound shot.
35mm? You sound like the kind of guy who gets paid to do this.
EDIT: It occurs to me some might ask: Why would anyone want to do this? Well, film is in resurgence. Not for whole projects. Typically, a hybrid approach where the majority is shot on digital. But for those shots with a lot of motion - action shots, or fast pans - where rolling shutter affects the image from a digital sensor, a few shots done on real film stock gets you global shutter on the cheap. A 200ft roll of 16mm processed and scanned to 4k is just a few hundred bucks. Who’s doing this? Music videos. Shorts. And school productions.
The goal here is to remove old reel to reel and 4 track tape recorders from the workflow. To record the sync signal from the camera and audio during filming directly to a digital recorder.
Film would be lab processed and scanned to a digital ProRes or DPX file at 23.997 fps. A known and synchronized framerate. Audio would have been recorded on a digital recorder, known and synchronized. But there would still be unsyncrhonized drift between video playback and audio…So, the pulse stream gives enough data to compress or stretch audio per frame to synchronize.
Think of it like using the sync pulses to snap audio to the grid. There’s already tools to do that. And there are already tools to stretch or compress audio. In this case the image frame rate would be about 41.6 milliseconds per frame. Audio drift somewhere ± 1 millisecond per frame. And a pulse recording from the camera as reference.
I understand if Ardour devs aren’t interested. But this is definitely possible to do. And Ardour already does most of it.
A little off topic, but why would you use drop rate from 24 fps? I have heard of a lot of people still using drop rates from 30 fps (even though it is no longer needed since there are no analog TV broadcasters still operating), but I don’t think I’ve ever heard of running 23.997 fps before.
The video file framerate should be exactly the same that you shot the film with, otherwise you get in trouble. If you are planning to using 23.976 fps you probably already are in trouble because of the poor NTSC standard Note that 23.976 fps and pulldown are only needed if you will broadcast your film on NTSC television. On file based platforms Youtube etc you can use 24, 25 fps and this will simplify your post workflow.
I suggest you use the old fashioned very very simple workflow. Just clap your hands or a clapperboard in the picture in the beginning (or end) of each scene. Since this is film you will be shooting only a couple of minutes max anyway. Then sync sound to video with the clap. Each scene should stay in sync for the duration of the scene with no problems as long as your cameras speed is stable and correct.
I used to be on a TV film crew in the 90’s We did not sync audio (Nagra / Dat) to camera (Arriflex 16 SR II) when shooting because there was no need. The camera and audio device could maintain the correct speed for the duration of a complete film roll (11 minutes). We did shoot at 25 frames per second since we are in Europe.
Test your post workflow before production so that you know it works.
That ARRI 16 you used had a crystal sync motor. Either modded from the original or sold that way by the 1990s. Which is why you didn’t run a cable from the camera to your sound recorder. And just slated without thinking about it. But in the old old days there were no quartz crystal oscillators! That’s like an early 1970s development. Think about the first quartz watches. And that technology didn’t make it to film production until the late 1970s.
Here are photos taken from my copy of The Filmmaker’s Handbook, from 1984, which I bought almost thirty-five years ago. In it is discussed the use of cable-sync double system sound. Which from the early 1930s on to the late 1960s, was the only way to synchronize film with sound.
There are still a lot of ARRI, Bolex, Beauliu, Eclair, Braun, etc 16mm and S8mm cameras out there which will only do cable sync. Because crystal sync was very expensive. And everyone in news who were using 16mm transitioned to video tape as fast as they could. And these cameras are built like tanks and easy to maintain. 70 year old cameras are still in use and available to rent. People mostly use these to shoot without sound and mix external narration and music in montage.
Anyway, it is possible to use a digital recorder and do traditional cable-sync double system sound with these old beasts. All it would take is software to correct audio drift from the recorded sync pulse.
I took a look at you pictures and wow what a wonderful old book
Would you care to tell us more about your project, it sounds very interesting. Why are you using old equipment for it, are you in it for the “film experience” ? I can totally get that, my wife just bought an old Pentax to shoot 35mm pictures and I’m into old computers (C64, Amiga, etc).
You seem to be using equipment before the 90’s and I don’t know much about those. However if you are using a digital recorder you can assume that it already runs at the correct speed, at least more correct than the camera.The only problem then is camera speed.
One can’t get very precise timing with old equipment even if it was in mint condition since the technology isn’t very accurate. I think it is quite enough to have each scene drift less than 1 film frame. Each time you start the camera you are anyway forced to manually sync picture and sound in post so it does not matter if the drift gets worse over a long period of time. As each scene is a new starting point for video and audio sync you don’t have to have precise timing over a long time, it is enough to have adequate timing over a scene (40 secs - 2 mins).
For most people to notice there is something wrong with lip sync the error needs to be more than 150 - 200 milliseconds. If video / audio drift is less than 40 milliseconds in 2 minutes then that is enough. Nobody (even the audio post guy) won’t notice it (I used to also work in audio post ).
What I’m trying to say is that maybe you don’t need pilot sync tone, but then again I don’t know anything about your project.
The problem with analog sync signals is that digital equipment don’t have any need for those and support might not be there. Maybe you explored this already and found some solutions. Not many people use old equipment, so the support is going to get even worse over time. One solution might be to use an analog audio recorder equipped with pilot capability and then digitize sound from the audio device as it is syncing itself to the recorded pilot tone. Then you would have a digitized sound that runs at the correct pilot tone pulse speed. As pilot would have been recorded from the camera then the camera -> audiorecorder -> digitized sound would all be synced to the “same master”.
All this difficulty is part of the “film experience” and much of it is the reason we moved away from film and analog audio. But hey, it is a very different experience and it was how it was done for almost a century
Precise timing for sync can be done on these old cameras with a cable sync system. Which is exactly how sound was sync’d before crystal oscillators were available. This problem has a software solution. Ardour seems to have pretty much everything under the hood to solve it. And that’s why I’m asking devs and the community here about it. That particular issue.
Why I (and a lot of other people) want this is irrelevant. It’s dependent on project needs. Though I did explain in a prior comment.
It’s a millisecond or less drift (about) per 41.6 millisecond long frame. So, I don’t think there would be significant or noticeable audio degradation. Tape decks would speed or slow tape transport by a tiny amount based on sync pulses. In any one frame drift is negligible. It’s across tens or hundreds of frames where drift builds up to become noticeable. This is a phase locking problem.
The video is a 24 fps stream locked by the NLE at 23.967fps or 24fps (depending on NLE settings). Video does not have variable rate (and yes, phones - but no, don’t go there).
I can’t speak for Ardour internals, but would a 2.5% timestretch per 41.6 milliseconds be noticeable?
How about a 2 millisecond per 41.6 millisecond time-stretch?
And BTW: speeding up or slowing down tape (very slightly) to maintain phase lock most certainly did adjust pitch. It’s not like those old tape decks used sophisticated pitch manipulation along with external phase lock. But it was so small nobody noticed.
I mean, in ancient times when the world was black and white, Hollywood feature films went out using this technology. Talkies would not have existed without it. (though in those days the recording was impressed on a wax cylinder)
So you have a list of time-positions that is evenly spaced at intervals of 1001/24 ms that correspond to frames of the scanned film.
And then you have “impulses” stream from an audio-track. Assuming you can find a tool to analyze and detect those impulses, you an assign a time-stamp to each impulse (relative to the start of the audio-file) and create a time mapping from -> to.
This mapping is then to be applied in revese to the 2nd audio channel. Time-stretch it to the rigid film framerate.
This last step could could be done with the rubberband commandline tool which allows to read a map file:
from rubberband --help
A map file consists of a series of lines each having two numbers separated
by a single space. These are source and target sample frame numbers for fixed
time points within the audio data, defining a varying stretch factor through
The pulse stream depends on the make and model of the camera. But typically it’s a simple tone generator, anywhere from 60Hz to 1000Hz depending on manufacturer. There’s an impulse output per frame. That’s sent to one of the audio channels and recorded as standard audio.The other channel (or channels) records say dialog on set. And based on this pulse rate, a drift fix against a standard frame rate is calculated per frame, and used to modify the audio channel(s) per frame.
You got it!
I will go look at the man page for rubberband. Thanks for that tip!