There is a specific deadness that creeps into someone’s eyes the moment you point a lens at them and ask them to speak. You’ve seen it in a hundred corporate videos. The words are fine. The lighting is fine. But nobody at home believes a syllable of it, and they couldn’t tell you why.
The usual fix is to put the subject in a chair, seat an interviewer just off-axis, and let them have a real conversation. It works. People warm up, they forget the camera, they say something true. The cost is that they’re looking slightly off-frame for the entire piece — past you, past the lens, at a person the audience can’t see. That’s the right choice for documentary. It’s the wrong choice when the whole point is that this person is talking directly to you.
The other usual fix is a teleprompter with a script on it. That solves the eyeline and creates a worse problem.
The reading tell
Anyone can spot read-aloud. It isn’t one obvious thing, it’s a stack of small ones. The eyes track horizontally in a way conversation never does. Blink rate drops. Emphasis lands on the wrong words because the speaker is decoding a line rather than meaning it. Breath comes at punctuation instead of at thoughts.
You can coach some of it out. You can widen the line spacing, slow the scroll, get someone comfortable enough that they paraphrase. But the underlying problem doesn’t go away: you’ve asked a person to perform words instead of say them, and most people who aren’t actors are not good at that.
We wanted the eyeline of a prompter and the behaviour of a conversation. So we kept the glass and threw away the script.
What we actually did
On the Colorado Connector campaign we filmed people in twelve communities along the Front Range corridor, from Fort Collins down to Pueblo. Local voices making the case for a passenger rail line through their own downtowns. Every one of them needed to look straight at the viewer, and not one of them was an actor.
So we ran a teleprompter with no text in it. Instead of a script, the glass fed a live image of the interviewer’s face. The subject stood at the lens and had an ordinary conversation with a person they could see, at the exact optical axis of the camera.
The result is that everything the prompter was costing us goes away. There’s no line to decode, so emphasis falls where the speaker means it. Blink rate is normal because they’re listening to someone. They pause where a thought ends rather than where a comma sits. And because a human face is genuinely engaging, they hold the eyeline without being reminded — the thing you normally have to nag people about is the thing they now want to do.
Building one
You need less than you’d think. Any beam-splitter prompter will do, because you’re not asking it to scroll anything — it just has to put a reflected image where the lens is.
- The prompter rig itself, mounted as normal on the matte box or rods.
- A laptop or tablet feeding the glass instead of prompter software.
- A camera pointed at your interviewer — a webcam is genuinely fine. This is a 6-inch reflection, not a deliverable.
- A way to get the interviewer’s audio to the subject. In the open they can just hear each other. In a noisy plaza, give the subject an earpiece.
Two things worth getting right
Mirror the image. Prompter glass flips what it reflects. If you don’t flip the source first, your interviewer appears reversed — which nobody consciously notices and everybody subconsciously finds slightly wrong.
Match the brightness to the scene. A laptop at full output in a shaded plaza will spill onto the glass and lift your blacks. Dim it until the reflection is comfortably visible to the subject and invisible to the sensor, then check on a monitor rather than by eye.
When not to use it
This is a tool for direct address. If you’re making something observational — someone at work, a process, a story told in the third person — the off-axis eyeline is correct and you should leave it alone. Reach for eye-direct when the subject is making a case to the audience: testimonials, advocacy, recruitment, anything where the viewer needs to feel spoken to rather than shown something.
It also asks something of your interviewer. They’re on camera in the sense that their face is driving the performance, so they need to actually listen and react. A bored interviewer produces a bored subject, and now you can see it in the eyes.
Across twelve towns it held up in every setting we put it in — brick sidewalks, an open plaza, a rail yard, a park in full sun. The whole corridor reads as one conversation, which is exactly what the campaign needed.