How to make a talking head video that doesn’t look like you’re reading

A talking head video is one person on camera, speaking to the viewer. It’s cheap to make, it puts a real face in front of your audience, and with some planning one afternoon of filming can cover a month of content. It’s also the format with the least to hide behind: a face, a voice, and whatever is behind them. Every weakness shows, and the one viewers catch first, often without knowing why, is someone obviously reading. This guide covers how we’d set one up for a client, and how to deliver from a script so it sounds like talking. (It’s about filming a real person, not AI avatars.)

What is a talking head video?

A talking head video is a shot of a person speaking directly to the camera, usually framed from the chest or shoulders up. The name comes from television, where it describes an expert or presenter shot in a tight frame, with only their head and shoulders visible.

You’ll see them everywhere:

• Explainers and tutorials
• Founder and team introductions on company websites
• Social content on LinkedIn, YouTube, TikTok and Instagram
• Course lessons and online teaching
• Updates to customers, investors or an internal team
• Sales videos sent to a single prospect

The format works because a face talking to you directly is how people are wired to receive information from someone they’re deciding whether to trust.

Framing: where to put the person

Shoot at eye level. Put the camera lens level with the presenter’s eyes. A camera below looks up their nose and reads as imposing; a camera above makes them look small. Eye level is neutral, and neutral is what you want.

Leave headroom, but not much. In a standard frame, the eyes sit about a third of the way down from the top. Too much space above the head makes the person look like they’re sinking out of the shot.

Choose a size and keep it. A medium close-up, from mid-chest up, is the default: close enough to read expression, wide enough to show hand movement. A close-up, just the head and shoulders, feels more intimate and suits a strong, confident delivery. Pick one for the main angle and change size only for emphasis.

Frame for where it will be watched. Filming for YouTube or a website, frame horizontally. Filming for Shorts, Reels or TikTok, frame vertically, and remember the platform’s own buttons and captions cover the top and bottom of the screen. Keep the face in the upper-middle of the frame, clear of both.

Mind the background. It should say something about the person or business, or say nothing at all. A tidy bookshelf, a workshop, a studio: all fine. A doorway with people walking past is not. Add some distance between the presenter and the wall behind them, so they don’t look pinned to it.

Eye line: the thing that gives reading away

Viewers can tell when someone is reading, even when they couldn’t explain how. It’s geometry. When your eyes land on something that isn’t the camera, even a few degrees away, every viewer clocks it.

That’s the problem with reading a script off a laptop screen below the webcam, or off a phone propped beside the camera. Your eyes are pointed at the words, not at the lens, and the audience notices.

The fix is to put the words where the lens is:

• On a laptop or webcam: keep the line you’re reading right at the top of the screen, as close to the camera as it goes.
• With a separate camera: place the screen with your script directly beside or below the lens, as close as you can get it, and keep the reading line at that edge.
• For the cleanest result: a beam splitter (the angled glass in a teleprompter hood) puts the script directly in front of the lens, so you’re looking straight down the barrel while reading.

In Miracue you can move the reading line to wherever your lens is: high on the screen if the camera sits above it, centered if you’re using a beam splitter. Same script, same delivery, and now you’re talking to someone.

Light

Face a window. Soft daylight from in front of the presenter is the best free light there is. Put them facing it, not with their back to it; a window behind the presenter turns them into a silhouette.

Add one key light if you have it. A single soft light slightly to one side of the camera, a little above eye level, shapes the face. That’s enough for most talking heads.

Watch for mixed color. Daylight from a window and a warm bulb in the room have different color temperatures, and mixing them puts orange on one side of the face and blue on the other. Turn the room lights off, or match them.

Avoid overhead light. Ceiling lights put shadows in the eye sockets. If you can’t turn them off, get a light at face height to fill them in.

Sound

Viewers forgive average pictures. They don’t forgive bad sound.

Use a microphone that isn’t the camera’s. A lavalier (clip-on) mic or a small shotgun mic just out of frame will sound dramatically better than a built-in microphone across the room.

Kill the room. Hard, empty rooms echo. Soft furnishings, a rug, curtains or even a coat rack out of shot all absorb reflections. A furnished room sounds better than a bare one.

Listen before you record. Fridges, air conditioning and traffic are easy to miss in person and impossible to miss in the edit. Switch off what you can.

Delivering from a script

This is where most talking heads succeed or fail.

Memorizing works for a thirty-second piece. For anything longer, or for a batch of videos in one day, it’s slow and fragile, and the effort of remembering shows on your face. Ad-libbing from bullet points works for confident presenters but tends to ramble. A script read well is usually the fastest route to a tight, accurate video.

The trick is making it not sound read:

Write it to be spoken. Short sentences. Contractions. The words you’d actually use out loud. Read it aloud once before you film, and rewrite any line you trip on.

Use a teleprompter, not a laptop screen. A teleprompter keeps your eyes at the lens and the words moving at the right speed. If you’ve never used one, our guide to reading a teleprompter naturally covers the technique.

Let it follow you, rather than the other way round. A fixed-speed prompter makes you keep pace with it, which is where the robotic delivery comes from. A voice activated teleprompter moves at the speed you actually talk and waits when you pause.

Do it in chunks. You don’t have to deliver three minutes in one take. Film a section at a time, and give yourself permission to go again on any line. The edit joins them.

Talk to one person. Picture someone specific who’ll watch it and talk to them. It’s the single most effective way to take the “presentation voice” out of your delivery.

Editing a talking head

Cut the gaps. Trim the pauses and false starts. Tight pacing is most of what makes a talking head feel professional.

Use cutaways to hide the joins. Every time you cut between two takes, the frame jumps. Covering the join with a cutaway (a shot of hands, the product, the place) hides it and adds interest. List the cutaways before the shoot so you don’t forget to film them; a shot list is made for exactly this.

Punch in for emphasis. If you shot in 4K and deliver in HD, you can crop in on a key line to change the size without a second camera.

Caption it. Much of social video is watched with the sound off. Burned-in captions keep those viewers.

Common mistakes

1. Reading off a screen below the camera. It’s the most visible tell there is.
2. Relying on the camera’s microphone. Poor sound loses viewers faster than poor pictures.
3. Sitting with a window behind you. You’ll be a silhouette.
4. Trying to do it in one take. Film in sections and join them.
5. A cluttered or busy background. It competes with your face for attention.

Batching: a month in an afternoon

The reason talking heads are so efficient is that the setup is the expensive part. Once the camera, light and microphone are placed, each additional video costs only the time to read it.

So batch them. Write six or eight scripts, set up once, and film them back to back, changing only your top between them if they’ll go out on different days. It’s how we’d run a content day for a client. In Miracue you can paste a whole shoot day’s worth of scripts in one go, and it splits them into separate scripts, ready to read in order.

Frequently asked questions