How to read a script and still look at the camera

There are only three ways to say written words to a camera and still appear to be talking to somebody. You can learn the lines. You can glance at a page and look back up. Or you can put the words in the one place your eyes are already pointing, which is next to the lens. The third is the only one that scales past a paragraph, and most of what follows is about doing it well, because doing it badly is why prompted video still looks prompted.

Why looking down is so obvious

A camera flattens everything except the eyes. Your audience cannot see the notes on your desk, so a downward glance has no explanation attached to it, and the brain fills in the gap with something unflattering. Reading, or worse, evasion. It takes about a quarter of a second and it lands every time.

The distance matters more than the direction. Eyes drifting a few degrees below the lens reads as thoughtful. Eyes travelling to a laptop screen sitting below and to the left reads as somebody following instructions. That is the gap you are trying to close, and every method below is a different way of closing it.

The four setups, and when each one is worth it

Learning the lines

Fine for thirty seconds, punishing after that. Memorising turns a reading job into a recall job, and recall is the first thing to go once the light is on. It also flattens delivery, because you end up listening to yourself retrieve the next clause instead of meaning the one you are saying. Keep it for an opening line you want to nail.

Notes beside the camera

Bullet points on a card taped next to the lens, big enough to read at a glance. Cheap, and genuinely good when you know the material and only need reminding of the running order. It falls apart the moment you need exact wording, a number you must get right, or a legal sentence that cannot drift.

A prompter in front of the glass

The beam-splitter rig, where text reflects off angled glass the camera shoots through. Perfect eyeline, because you are reading through the lens rather than beside it. Worth it if you record to a proper camera on a tripod several times a week. For a laptop, a webcam and a Tuesday afternoon, it is a lot of hardware to own and align.

The script on the screen you are recording with

Words at the top of the same screen the camera sits above, so the travel is a couple of degrees. This is what most people should use, and it needs no hardware at all. Its whole quality depends on one thing, which is the next section.

The part that gives it away is pace, not eyeline

Get a prompter, set it scrolling, and you can still look exactly like somebody reading. Fixed-speed scrolling is why.

A constant scroll has no idea what you are doing. Set it fast enough to feel natural and you spend the take chasing it, clipping the ends of sentences to keep up. Set it slow and you arrive at the bottom of the visible text with nothing to say, so you slow your speech to match a machine, and the pauses land in the wrong places. Either way the rhythm belongs to the scroll rather than the sentence, and that is what an audience hears.

The fix is a prompter that moves when you move. Speed up because you are enjoying a passage and the words keep up. Stop to think and it waits. Deliver a sentence in your own words rather than the written ones and it finds you again when you rejoin. You stop managing the text and go back to talking, and that is most of the difference between prompted and natural.

Setting it up so it works the first time

• Put the reading line as close to the lens as the software allows, and keep the area right at the top of the screen clear of controls.
• Sit further back than feels normal. An arm's length flattens the eye travel, and it stops the wide-angle distortion that makes faces look odd up close.
• Make the text bigger than you think you need. If you are squinting, your eyes narrow on camera and it reads as strain.
• Keep lines short. Long lines force your eyes to sweep sideways, which is as visible as looking down.
• Check the running time before you record rather than after. Knowing a script is nine minutes changes how you write it.

Write it for the mouth, not the page

Half of looking natural happens before you record, in the script itself. Written English and spoken English are different languages, and a sentence built for the eye is hard to say out loud without sounding like it was built for the eye.

Read every line aloud as you write it. If you run out of breath, split it. Use contractions, because you use them when you talk. Cut the clause that qualifies the clause. Where you would naturally say "so" or "anyway" or "here is the thing", leave it in, because those are the joins that tell a listener how the parts relate.

One test settles most of it. Say the sentence to a friend, from memory, in your own words. Whatever came out is closer to the script you want than what you originally typed.

What good looks like

You are reading every word and nobody watching would guess. Your eyes stay on the lens, your pace is yours, and when you stumble you keep going, because a small stumble reads as a person thinking. If you want the longer argument for that last one, it is in why the sixth take is rarely better than the first.

LensFloat does the fourth setup above. You paste in your script, it sits at the lens and keeps pace with you rather than with a timer, and it records while it does it. If a deck or a whiteboard belongs in the same video, those are on the same stage instead of in another window.