Last month I cut an episode where the host and the guest never once looked at each other. Not in a rude way. The host was reading the guest's face on a second monitor off to the left, and the guest was watching a browser window parked at the bottom of the screen. On camera it read as two people talking to someone standing just outside the frame. The conversation was great. The video felt like a hostage exchange.
That is the remote guest problem in one sentence. The audio can be perfect and the episode can still feel wrong, because eye contact is the thing your audience checks first and cannot name.
I have cut hundreds of remote episodes now, and I have stopped treating this as a camera issue. It is a setup issue, a hosting issue, and to a smaller extent an editing issue. In that order.
Why the eyes matter more than the lens
People forgive soft focus. They forgive a slightly noisy room. They do not forgive a person who appears to be avoiding them.
We read gaze faster than we read anything else in a frame. A viewer decides whether two people are actually connecting inside about three seconds, and that read happens before they have processed a single word. When the gaze is off, the brain files the whole thing under "performance" instead of "conversation". Everything after that gets a little less trust.
This is why the same guest can sound brilliant on the audio feed and look checked out on the video feed. Nothing changed except where their pupils were pointing.
A remote guest is not looking away because they are bored. They are looking away because you put the interesting thing somewhere other than the camera.
That is the whole fix, really. Put the interesting thing near the lens.
The four setups I see, ranked
Almost every remote guest I receive footage from falls into one of four camps.
- Laptop webcam, guest window in the middle of the screen. Bad, and the most common. The camera sits above the screen, the face sits below it, and the guest spends the entire hour looking at a point three inches under the lens. Reads as looking down at their notes.
- External camera on a tripod, guest window on a second monitor. Worse. Now the gaze is off to the side by fifteen or twenty degrees, which our brains read as "talking to someone else in the room".
- External camera, guest window shrunk and dragged directly under the lens. Good. The offset drops to a few degrees and stops registering.
- Camera behind a teleprompter glass with the call feed on it. Best, and almost nobody does it, because it costs money and takes a shelf.
The gap between the second setup and the third one is free. It is dragging a window and resizing it. I have watched that single change turn an episode from stiff to warm with no other adjustment.
What I ask guests to do before we record
I do not send guests a technical document. Guests will not read a technical document. I send three lines, and I ask the host to paste them into the calendar invite so they arrive at the right moment.
- Shrink the call window to about a third of the screen and drag it right under your camera.
- Close everything else, especially anything with a notification.
- If you have notes, put them on paper next to the camera, not on the screen.
That third one matters more than people expect. A guest reading notes on screen looks like a guest reading notes on screen, and no amount of cutting hides it. Paper held slightly to the side of the lens produces a glance, which reads as thinking. A screen produces a scan, which reads as reciting.
I go further on this in the case for a short call before the interview, because the pre-interview is where you actually get to see their setup and fix it while it is still cheap to fix.
The host is the bigger offender
Here is the part hosts do not like hearing. In most of the footage I get, the guest's gaze is fine and the host's gaze is a mess.
Guests are on one screen doing one thing. Hosts are running the recording software, watching a question doc, checking the levels, watching the clock, and sometimes messaging a producer. Every one of those is a place for their eyes to go, and every one of them is somewhere other than the lens.
If you host, do this: get your questions down to six bullets on an index card, put the card just beside the lens, and close the recording software window once the record button is green. You do not need to watch the meter. If the meter is going to fail, it will fail whether you are staring at it or not.
The other version of this fix is structural. When there are two of you, one person can watch the levels and the clock while the other just talks. That is one of the quiet arguments in favour of running the show with a second person, even a part time one. It buys the host their eyes back.
What editing can and cannot fix
Let me be honest about my own leverage here, because there is a lot of nonsense sold about this.
I cannot rotate someone's eyes. Nobody can, not without the result looking like a deepfake, and I would not ship that anyway.
What I can do:
- Cut to the speaker less often when the gaze is bad. If a guest is looking hard off axis, holding on them in a full screen makes it obvious. Sitting in a side by side or on the host reduces the sting.
- Reframe. Pushing in and shifting the frame so the guest's gaze leads into empty space rather than into the edge of frame makes the offset look intentional. It is an old cinematography trick and it works.
- Cut on the look. If the guest glances back toward the lens on a strong line, that is my cut point. The audience sees the connected version of them.
- Use the reaction. A host nodding at a good line covers a guest's worst gaze moment for free.
None of that is invisible work. It just costs time, and it is time spent covering for something that a thirty second setup change would have prevented. Every human editor on our team watches the whole episode, so we catch these moments, but I would rather spend those hours on pacing than on damage control.
There is a related trap on the other side of this, which is that a guest who is uncomfortable on camera tends to talk longer to fill the silence. If you are also fighting answers that run three minutes too long, the root cause is often the same discomfort.
The interruption problem is a gaze problem too
Remote conversation has no shared room, so it has no shared timing. You cannot lean forward. You cannot catch someone's eye. Latency eats the half second where a normal interruption lives.
The result is one of two failure modes. Either the host never interrupts and the episode sags, or both people talk over each other constantly and I spend an afternoon untangling it.
The workaround that actually holds up is a physical signal. A raised hand into frame. Not a wave, just a hand. The guest sees it in their window, stops, and nobody has to talk over anybody. It looks slightly odd in isolation and completely normal in context, and it saves an hour of editing. If you want the softer verbal versions, I wrote about stopping a guest without being rude separately.
Where camera count actually helps
This is where remote and in person diverge, and where I think people spend money in the wrong place.
In an in person show, more cameras buy you rhythm. Our multi-cam editing work leans on that hard, because a second angle gives me somewhere to go on every beat.
In a remote show, the guest gives you exactly one angle no matter what you do. Adding cameras on the host end helps a little, mostly for reaction shots and for hiding the guest's weakest moments. But it does not solve the connection problem, because the connection problem lives on the guest's side of the call and you do not control it.
So my advice for remote first shows is unglamorous. Spend on the host's framing and lighting, spend on the guest's brief, and do not buy a third camera hoping it fixes intimacy. Two clean angles and a guest who is looking at the lens beat four angles and a guest who is looking at their taskbar.
Episode length interacts with this too. A gaze problem that is tolerable at thirty minutes becomes exhausting at ninety, which is one more reason to think about how long the episode should actually be before you record it rather than after.
The checklist I would tape to your monitor
If you take one thing from this, take the list.
- Call window shrunk and parked directly under the lens, both sides.
- Camera at eye height. Not below, never below.
- Notes on paper, beside the lens.
- Notifications off, second monitor dark.
- Host closes the recording software once it is rolling.
- Agree the raised hand signal in the first minute, before the record button.
- Both sides record locally, and everything gets sent over in one clearly named folder so nothing goes missing.
That last one is a different article, and it is the one that bites hardest when it goes wrong. If your remote workflow does not have a plan for getting files across without breaking them, the eye contact will be the least of your problems.
The remote episode will never feel like two people in a room. I have stopped trying to make it feel that way, and I think chasing that is why a lot of remote shows look overproduced.
What a remote episode can be is honest. Two people, clearly present, clearly listening, clearly looking at each other as best the technology allows. That is achievable in about ninety seconds of setup, and it is worth more than any grade or graphic package I could put on top.
If your last few episodes felt flat and you cannot say why, watch thirty seconds on mute and only watch the eyes. You will probably find it. And if you want a second pair of eyes on it, you know where to reach me.