Recording a video podcast people actually finish: the room, the mics and three cameras

People will watch a podcast that looks ordinary. They will not finish one that sounds bad.

That single asymmetry should decide most of the budget, and it usually does not — the money goes to cameras, the audio is whatever the room and a pair of mics produced, and the episode loses its audience in the first ninety seconds for a reason nobody in the room could name.

Here is what actually makes a recording watchable, in the order it matters.

The room decides your audio before a microphone is chosen

The hollow, distant, inside a bathroom quality on amateur podcasts is not a microphone problem. It is reverb — sound bouncing off hard parallel surfaces and arriving at the mic a second time, slightly late.

No microphone rejects it and no plugin removes it cleanly. It has to be dealt with in the room.

What a good room has:

• Soft mass on the reflective surfaces. Absorption panels on the first reflection points, a rug on a hard floor, heavy curtains across glass. Books on shelves work genuinely well.
• No parallel bare walls. Two hard walls facing each other produce flutter echo — the metallic ring you hear when you clap in an empty room.
• A short decay. For speech, a reverberation time under roughly 0.3 seconds is the target. Past that, intelligibility drops and every consonant smears.
• Silence underneath. Air conditioning, a ceiling fan, a fridge, traffic. The noise floor is inaudible while you are in the room and unmistakable in headphones.

The test costs nothing: stand where the guest will sit and clap once. If the sound rings on after your hands part, the room needs treating before it needs equipping.

One microphone per person, close, and always

Not a shared mic in the middle of the table. Not the camera’s mic. One per speaker, positioned close.

Distance is the whole game. A microphone six inches from a mouth records mostly the person. The same microphone three feet away records mostly the room, and the room is the thing you just spent effort suppressing. Halving the distance to a source is worth more than any upgrade in the microphone itself.

Two practical rules that prevent most disasters:

• Record each mic to its own track. One guest who talks quietly and one who booms are fixable on separate tracks and permanently baked together on a single mixed one.
• Run a safety recording. A second recorder, or the camera’s own audio, running in parallel. It is unglamorous, and it is what stands between a technical fault and a lost interview that a guest will not sit for twice.

Three cameras, which is the sweet spot

One camera is a static talking head and gives an editor nothing — no way to cut, no way to remove a stumble, no change of pace across forty minutes.

Five cameras is a switching job that needs someone whose only role is watching them.

Three is where it settles, and the arrangement is standard because it works:

1. A wide holding both people. This is the establishing frame and the safety net — it is always usable, so every cut has somewhere to go.
2. A tight on the host.
3. A tight on the guest.

That gives an editor a cut on every exchange, a reaction shot whenever the listener’s face is better than the speaker’s, and a wide to fall back on. Add a fourth only for a third participant.

Put them at eye level. A camera above the subject looks down on them, and a camera below looks up their nose. Eye level is the neutral, natural framing, and it is free.

Lighting, briefly

Podcast lighting has one job: make faces look like faces, consistently, for an hour.

A key light slightly off-axis, something soft filling the shadow side so it does not go black, and a little separation from the background so the subject is not fused to the wall behind them. That is the whole recipe.

The failure to avoid is mixing colour temperatures — a warm lamp on one side, daylight from a window on the other, and a guest with one orange cheek and one blue one. Close the curtain or match the sources.

The decisions to make before anyone arrives

• How long is an episode, really? A 40-minute conversation is a different production to a two-hour one: batteries, cards, water, and how long a guest will genuinely stay sharp.
• Which cuts do you need? The full episode, plus vertical clips for reels, is two different framings. Vertical clips need headroom to crop into, and that has to be framed during the shoot — it cannot be added afterwards.
• Who is watching the sound while you talk? Hosting and monitoring at once does not work. Someone in headphones, whose only job is listening, catches the fault at minute three instead of at minute fifty.
• How does it get off the cards? Two copies before anyone leaves the building.

Batch, because the setup is the expensive part

Lighting, framing and levelling a three-camera podcast takes a real chunk of a session. Doing it once and recording four episodes costs very little more than doing it once and recording one.

Every serious podcast we work with records in blocks. It is the single biggest saving available, it makes the look consistent across a season, and it removes the weekly scramble that kills most shows by episode six.

What we run

Our floor in Bhubaneswar is set up for exactly this: a treated room, mics per person on separate tracks, three cameras at eye level, and lighting that is already rigged when you arrive — so a session starts with the conversation rather than an hour of setup. It is available with a crew or as a room to hire.

Either way, the order of priorities is the same one above. Fix the room, get a mic on each person, give the editor three angles. Everything after that is refinement.