One take, three problems: the weaver interview
An NGO has one hour with a 78-year-old weaver who has agreed to be interviewed once and will not repeat himself. The room is a community hall with pale green walls, two LED tube fixtures on the ceiling, a tungsten bulb in the side corridor and a large window onto an overcast street. The generator is 40 m away behind a wall. The camera is a mirrorless body with a 35 mm lens; the operator notices horizontal banding rolling through the monitor, a green cast on the subject's left cheek, and a steady hum in the headphones. There is no time for a second take and no budget for a lighting truck. The sound recordist has a boom, one lavalier and a spare card. The interview will run 25 minutes; the final edit needs 3 minutes plus clean cutaways.
1. Diagnose the three symptoms and rank the fixes in the order you would apply them on set. Explain why that order and not another.
Read the worked answer
1. Rolling horizontal bands. Cause: the LED tubes are driven by a switching driver that is not flicker-free, and the camera shutter is not synchronised to the 50 Hz mains. Test in five seconds: shoot a clip, play it back at quarter speed - if the bands move, it is the shutter; if they sit still and pulse, it is the driver. Fix: set shutter to 1/50 s and shoot a test. If bands persist, kill the two LED tubes and relight with the tungsten bulb plus bounced window light, or swap in a known flicker-free panel. Never fix flicker by dimming an LED - dimming makes it worse.
2. Hum. Confirm it is the generator, not the camera or a ground loop, by unplugging the headphones from the camera and listening to the room. Fix: boom to within 30–40 cm above frame line, lav as an isolated backup track, all cables dressed away from mains runs, and a 60-second room tone recorded in dead silence.
3. Green cast on the left cheek. Three possible sources and you must separate them: the pale green wall bouncing light back onto the shaded side of the face; the LED tubes' green spike; or white balance set to daylight while the tube is a different colour temperature. Look at the white shirt and the wall in the monitor - if the shirt is neutral and the wall is green, it is real bounce. Fix: move the chair 1.5–2 m off the wall, hang a black cloth on the wall behind the subject's shoulder, and set white balance manually with a grey card facing the camera.
Order: banding, then hum, then colour. Banding is the only one that is expensive or impossible to fix in post. Hum is partly fixable with a notch filter. Colour is the cheapest thing in the grade - so it goes last, even though it is the most visible complaint.
2. State the camera and audio settings you would lock in before the first question, and explain what each choice is protecting against.
Read the worked answer
Camera
- 25 fps - the delivery frame rate in India; it also makes 1/50 s a true 180-degree shutter.
- 1/50 s shutter - the anti-flicker setting for 50 Hz mains. Locking it removes the banding problem at the source.
- Manual white balance, set off a grey card in the key light, not auto - auto WB will chase the overcast window and shift skin tone mid-answer.
- Fixed ISO and aperture, with ND on the window if the glass blows out. Choose exposure for the face and accept a clipped window; a blown window reads as intention, a dark face reads as error.
- Flat but not log profile - log needs a grade that a 3-minute NGO edit may not budget for.
- Manual focus, peaking on, locked - the subject will lean forward as he talks.
Audio
- Boom primary, lavalier as a separate isolated backup track, 24-bit / 48 kHz, peaks around −12 dBFS, limiter off unless you are certain it will not engage.
- Headphones on for every second of the take, including the pauses.
- 60 seconds of room tone with the crew silent - the editor cannot build a noise profile without it and cannot fake one.
- A 30-second recording of the generator alone with the subject out of the room, so post has a clean sample of the exact noise to subtract.
Every one of these is a decision that costs seconds now and hours later. On a single-take interview, the settings are the production.
3. The generator cannot be switched off and the weaver will not repeat himself. Describe your plan B for the sound and exactly what you tell the editor.
Read the worked answer
Plan B on set, in order:
- Ask for two minutes before the interview starts to record the generator alone and the room tone. This is the single most valuable thing you can walk away with, and it costs the subject nothing.
- Get the boom as close as politeness and frame allow; a 35 mm lens on a seated subject means 30–40 cm is achievable.
- Place the lavalier under the collar on the chest, not on fabric that rustles, and record it to a second track - two imperfect recordings that differ in noise are far better than one.
- Ask the subject to pause for five seconds after each answer rather than speaking continuously; the pauses give the editor room to cut without touching words.
- If the hum is audible in the headphones at a level you know will hurt dialogue, say so to the producer immediately and ask for the generator to be moved behind the second wall or baffled with plywood. Silent acceptance is the failure here, not the generator.
What you tell the editor, in writing, with the drive:
- The hum is constant, not intermittent, at mains frequency and harmonics, with a broadband component from the cooling fan.
- There is a clean generator sample and 60 seconds of room tone at the head of the card.
- Suggested chain: noise print from the sample, subtractive reduction limited to 6–10 dB, notch at 50, 100 and 150 Hz, high-pass at 90 Hz, gentle de-ess, then compression.
- The banding was solved in camera, so no de-flicker needed - do not waste an hour on it.
- The green cast is wall bounce on the left cheek; the grade will need a secondary on that side, not a global shift.