Eye contact on camera is geometry before it is acting. The angle between "looking at your script" and "looking into the lens" is what viewers see, and you shrink it with placement and distance — where the words sit relative to the lens, and how far you stand from both. Get those two right and even a plainly visible script reads as eye contact; get them wrong and no amount of discipline saves the take.
This guide covers the geometry, the placements available to you in rough order of cost, and the software settings that close the last few degrees.
Why your eyes give you away
The camera magnifies eye movement. A glance that feels imperceptible to you — a flick down to catch a line, a drift sideways to a second screen — is plainly visible in the footage, because the viewer is watching your face at a size and steadiness no live conversation provides. And the direction of the drift matters: downward glances read as reading or evasion, sideways glances read as distraction, while small vertical movement near the lens mostly passes as thought.
The instinct is to fix this with willpower — hold the lens, glance faster. It does not work, because the problem is angular, not moral. If the script sits thirty degrees from the lens, your eyes travel thirty degrees to read it, and the viewer sees every one of them. Move the script to five degrees from the lens and the same glance becomes invisible. That is the whole trick, and every placement below is just a way of buying smaller angles.
Distance does half the work for free
Before moving the script, move yourself. The angle your eyes travel depends on the gap between script and lens and on your distance from them: the same ten centimetres of gap is a large angle at arm's length and a small one at three metres. This is why a phone beside a camera works across a room and fails on a desk, and it is the cheapest improvement available — step back, let the framing loosen, crop in afterwards if you need to.
At one to three metres — the range most solo rigs end up at — text size becomes the constraint instead. The words have to land in glances and peripheral vision, which means bigger and bolder than any default. If you are squinting, the viewer can see that too.
The placements, honestly compared
Script under the lens. The strongest free option. Vertical eye movement is the kind viewers forgive, so a script directly below the camera — a narrow window at the top of a screen with a webcam above it, or a phone mounted just under a camera — keeps your gaze within a few degrees of the lens on the most forgiving axis. The desk setup guide is this placement in full.
Script beside the lens. Workable at distance, weak up close, because horizontal movement is the kind that reads as reading. If sideways is what your rig allows, distance matters even more, and the script should be as tight against the camera as physically possible.
Words on the lens axis. The complete solution is a beam-splitter rig: half-mirrored glass over the lens, so the script floats exactly where the camera is and reading is eye contact. It costs money or a weekend of building, and it brings its own rules — the glass eats a third to half the light, so text wants to be larger and bolder still, and the mirror flips everything, which is what mirror mode is for.
The software's share of the angle
Three settings close the remaining degrees, whatever the placement.
Active line position. The line you are reading is the only line your eyes need, so put it nearest the lens: at the top of the script area for an under-lens rig, wherever the lens axis falls on a beam-splitter. TellyPrompter lets you set where the active line sits rather than assuming the middle of the screen.
Look-ahead. Enough upcoming text to feel safe, not so much that your eyes wander down the page hunting the future. A shallow band of script keeps the eyes home.
A scroll that does not send your eyes hunting. The most common eye-line break is not a glance at the script — it is losing your place, and the visible search that follows. On a fixed scroll, every pause and stumble invites one. Voice-following removes the class of problem: the highlight sits on the word you are saying and waits when you stop, so there is nothing to find, and a glance at the script lands exactly where you are. Speech recognition runs on the device and the audio is never uploaded. The honest limit: it relies on the browser's speech recognition, which Firefox does not provide — the other scroll modes work there.
Practice the one habit that remains
With the geometry handled, one habit finishes the job: talk to the lens, not at it. Pick the small dark circle and treat it as a person's eyes — hold it the way you would hold a friend's gaze while making a point, which includes the natural tiny movements of a real face, not a fixed stare. If the talking head format is your home ground, that guide covers the delivery half; and if the read itself is what pulls your eyes loose, start with reading naturally. The geometry buys you the freedom; the talking is still yours to do.