How do you look at the camera while reading a script?

Short answer: move the words to the lens. Reading is not the problem: where you read from is. If the text sits in the middle of the screen, your eyes sit below the lens, and to whoever is watching that looks like you are avoiding them.

Why practice does not fix it

The gap between the lens and mid-screen text is a few centimetres at arm's length, which is enough for a viewer to read as broken eye contact. It is geometry, not skill. Presenters who "just memorise it" are solving the same problem by removing the text entirely, at the cost of sounding recited.

How far is "a few centimetres"?

Far enough to matter. The angle your eyes drop is set by two numbers: how far the line you are reading sits below the lens, and how far the screen is from your eyes. This table is a calculation (the angle is the arctangent of offset divided by distance), not a measurement:

Text below the lensPhone at 30 cmPhone at 40 cmLaptop at 60 cm
1 cm1.9°1.4°1.0°
2 cm3.8°2.9°1.9°
3 cm5.7°4.3°2.9°
5 cm9.5°7.1°4.8°
7 cm13.1°9.9°6.7°
10 cm18.4°14.0°9.5°

Vision research gives those numbers a scale. Studies of video gaze found that viewers stop seeing eye contact when the speaker looks more than about 5.5° below the camera (about 4.5° above it), and a 2024 study found that looking about 2° below the lens reads as eye contact even better than looking straight into it. On a phone, the middle of the screen is roughly 6 to 8 cm below the front camera: about 8° to 15° at a normal distance, well past that limit. A line right under the camera, 1 to 3 cm below it, stays within that limit or right at it.

Try your own setup: eye contact calculator.

Sources: Gao, Jayakumar, Maniglia, Curless, Kemelmacher-Shlizerman, Seitz and Seitz, Don't Look at the Camera: Achieving Perceived Eye Contact, Journal of Vision, 2024 (which also summarises the earlier threshold studies); Chen, Leveraging the Asymmetric Sensitivity of Eye Contact for Videoconference, CHI 2002, which found viewers are less sensitive to gaze below the camera than above or beside it.

The three ways people solve it

  1. Hardware teleprompter. A beam-splitter puts the text optically on the lens axis. It works perfectly and costs money, setup and desk space.
  2. Memorise and improvise. Free, and it is why so much on-camera speech sounds either recited or lost.
  3. Put the script at the lens in software. A reading band anchored under the front camera instead of mid-screen. Your eyes stay within the few degrees that read as eye contact.

What else matters once the text is in the right place

How EyePrompt does it

EyePrompt anchors the reading band directly under the front camera and scrolls it with Smart Follow, which uses on-device speech recognition to follow your voice and hold when you stop. It runs on iPhone, iPad and Mac from one purchase, works entirely offline, and uploads nothing. It is free to start, and on the free plan the prompter has no time limit.

Download on the App Store

What to look for in a teleprompter app

Questions people ask

How do you look at the camera while reading a script?

Move the words to the lens. Text in the middle of the screen puts your eyes below the camera, which viewers read as avoided eye contact. A beam-splitter teleprompter solves it optically; a software prompter solves it by anchoring the reading band directly under the front camera.

Why does practice not fix reading off-camera?

It is geometry rather than skill. The gap between the lens and mid-screen text is enough for a viewer to read as broken eye contact at normal distances, no matter how well rehearsed the delivery is.

How far below the camera can you look and still make eye contact?

Studies of video gaze put the limit at about 5.5 degrees below the camera, and a 2024 study found that about 2 degrees below the lens reads as eye contact best. On a phone at 40 cm, text 3 cm below the lens is about 4 degrees; the middle of the screen, 6 to 8 cm below, is about 8.5 to 11 degrees.

Does the scroll speed matter?

Yes. A fixed-speed scroll makes the reader chase the text, which produces a recognisable tell. A prompter that follows the voice keeps the current line under the reader's eyes when they slow down or pause.

Google lets you pick the sources you want to see more of. You can add EyePrompt as a preferred source on Google.