Harpo’s and Chico’s musical performances were almost certainly performed live in 1929–1930, and appear to have been continued that way into their M-G-M period in the mid-1930s onward. I play piano, and have watched Chico’s playing closely to detect if he was performing a pre-recorded playback, but his playing is note perfect. It would have been easier to just record him live than have to do multiple takes until he got every note synched to the playback.
Most vocal numbers in early musicals, 1927–1932, were recorded live. But there were exceptions, usually for dance numbers. For example, a musical number in The Hollywood Revue of 1929 that involved overhead photography of dancers. And the number “The Wedding of the Painted Doll” in The Broadway Melody (1929) was refilmed to a sound playback when the studio was unhappy with the original staging. But the other numbers, such as this version of the title song, were filmed live.
There were two competing sound recording systems in Hollywood in the period 1927–1930, sound-on-disk (Vitaphone) and sound-on-film (Fox Movietone and RCA Photophone). The only studios that used Vitaphone were Warner Bros. and its sister studio First National. Sound-on-film eventually won out, and was how all soundtracks were recorded from 1930 until the studios converted to recording on magnetic tape in 1950.
When recording on the set with sound-on-film, the signal from the boom microphone(s) went over a line out of the soundstage, and across the studio lot like a telephone line to the sound recording building. In there, a battery of amplifiers and sound recording cameras converted the audio signal to a light signal, and recorded the optical soundtrack onto film. An on-the-set sound recordist remotely controlled the starting and stopping of the sound recording camera, and adjusted the signal level to keep it within parameters as scenes were being shot. The dialog track would later be mixed with a sound effects track and a music track to make the master soundtrack.
The Miracle of Sound, a 1940 M-G-M short subject on this process. (Be patient, it takes a while to download.)
1888–1925: Acoustic recordings. The sound went into a large recording horn, which narrowed to a cutting stylus attached to a diaphragm. The needle would cut through wax or foil on a rotating cylinder or disk. More here.
1925–1949: The sound went through an electric condenser microphone and amplifier to the cutting stylus, which was controlled by electromagnetic signals, which cut a groove into the wax master. More here.
1949–1980s: The sound went through an electric condenser microphone and amplifier, and then was recorded in an analog signal on magnetic tape.
Of course, if sound and picture were recorded on separate media, it wouldn’t take a genius to figure out that they could play back a perfect track and recreate the motion.
And unlike today’s lip-sync, they could sing, plink, or drum all they wanted out loud when working against the sound track. By the fifth take, and having done a song a hundred times on stage, I’m sure Harpo or anyone could play in harmony with himself well enough to fool the camera.
I bet the main reason for recording most film sound live was the cost of actually faking an entire soundtrack with studio recording and looping. It’s bad enough with today’s technology, imagine in the days before magnetic tape and instant playback. Also in the days when they didn’t have the financial luxury to churn out film or soundtrack discs/cylinders all day long, and then spend several days in the edit room, to produce 5 minutes of a film.
From everything I’ve ever heard by that time Harpo was an extremely accomplished harpist so it wouldn’t surpise me if he could pull it off with just a few takes. Also notice how intense he is playing his harp in those scenes so I would vote against his music being recorded elsewhere.
The problem comes in with the synching. Unless the camera and audio device are running at precisely the same relative speed, synch is lost. Even if it’s off by one frame, people will notice it.
That’s OK for a one-take performance. If it’s flubbed, just do it again. If you want multiple angles, then the actors can perform the routine for each setup. Now you have to match the editing to the music, and also ensure that the audio levels are the same for each shot. With a master audio track, the audio levels are the same all the way through, and editing would be a little easier.
Huh? If the sound and picture are recorded on different devices to start with (as mentioned earlier), you still have the synch problem. If you can synch the sound with the original picture, you can play it back during the retake and synch that retake to the recording. Again, anyone working with the equipment could probably design a device for playing rather than recording a sound track in synch with a picture. The trick would be getting the clacker guy to clack-sync.
If you are doing multiple angle shots, a master recording is the best way to ensure sound quality and volume consistency.
I suspect the problem is that editing with primitive equipment was extremely time-consuming, and this is the reason why such tricks were rarely done.
Someday I’ll have to watch those movies again and see how much is one-takes. (“Russian Ark” is famous for being one 70-minute take). Like the director though, I probably couldn’t concentrate. I think it was Night at the Opera, where Harpo(?) is climbing the backdrop as fast as it’s falling, where I could not stop from rolling around on the floor laughing and trying to catch my breath.
Also, we are so used to quick cuts to different point-of-view as a normal part of cinema, that we gloss over the fact that something could be multiple takes spliced together.
The problem is that motor speeds can vary. That’s why they’re known as ‘wild’ motors. You need a way to synch the camera and the recorder. As I mentioned earlier, this was done with a signal. A tone generator is put on the camera shaft. It generates a 60 Hz tone that is recorded on the recorder. With the ‘clack’ from the slate, the signal can be used to synch the audio to the image, even when the motor speeds varied during filming.
I forgot to say this earlier, but the cable-synch is more of a portable camera thing. The studios used cameras and recorders plugged into the mains. Since the mains are AC, they were a reliable source for a 60 Hz signal.
1930s image/audio synch isn’t really my area of expertise though, and my cameras have crystal motors.
RCA advertisement in Life magazine from 1939 explains the sound recording process for motion pictures. I like how the editors tried to sex-up the article’s title.
(Click the little magnifying glass above the window to enlarge the article.)
The dynamic range probably wasn’t a choice. The sound engineers were probably going to the limit of image resolution for an optical sound track moving at 18 inches per second. A magnetic tape moving at that speed is a whole 'nother ball game.
In the days before sound, 16 frames per second was sufficient to give the illusion of motion. Hand-cranked cameras obviously depended on the cameraman for its speed, so actual filming speeds were variable – as were projection speeds. With the advent of ‘talkies’ in the '20s, the films had to be shot at a constant speed for sound reproduction. It was found that 24 frames per second was the slowest speed at which the sound was acceptable. Why choose the slowest speed? Because film is expensive. Slowest possible speed = less film used = money saved.
Super-8 sound film, introduced in 1973 (super-8 silent film came out in 1965) uses a magnetic strip to record the sound. Walloon says it’s 'a whole ‘nother ball game’, which I take to mean that magnetic tape is capable of acceptable sound quality at 18 frames per second. (And it’s OK at 18 fps. Better at 24, though.) Regular 8mm film cameras typically ran at 16 fps, as did 16mm cameras for amateur use. (Most of the 16 mm footage I’ve seen looks like it was actually shot at the industry standard of 24 fps.) The standard speed for super-8 was 18 fps, which meant a 50-foot cartridge would last three minutes and 20 seconds. Many super-8 cameras, especially sound-capable ones, also offered 24 fps (two minutes and 30 seconds per roll).