The Elgato Prompter decides how your script looks; the software you run on it decides how your script reads. The device is, at heart, a beam-splitter with a built-in screen that your computer treats as a display — which means you aren't limited to the software in the box, and the choice of what to put on that screen matters more to your delivery than any hardware spec.
This piece is about that choice: the three ways prompter software can scroll, what each one asks of you mid-take, and how to think about browser-based prompters, offline shooting and the cases where a paid desktop suite genuinely is the better answer.
The hardware in one paragraph
The Elgato Prompter packages the classic beam-splitter design — angled half-mirrored glass with the camera shooting through it — around its own small screen, and connects to your computer over USB, where it shows up as an extra display. Elgato ships companion software that can put a script, your camera feed or other content on that screen. What the hardware decides: your eye line (excellent, since the script sits directly over the lens), your mounting options, and the physical size of the reading surface. What it does not decide: how the text moves. Anything you can show on a display, you can prompt from — including a browser tab — so the scrolling behaviour is entirely a software question.
That's worth dwelling on, because most reviews of the device spend their word count on build quality and lens compatibility and a paragraph on software. Yet the software is the half you'll be negotiating with mid-take.
The three ways prompter software can scroll
Every prompter app, whatever its badge, moves text in one of three ways.
Constant scroll. You set a speed — words per minute, or a slider — and the text moves at that speed from the moment you start. This is the traditional mode, and the one most prompter software defaults to.
Sound-level scroll. The halfway house: the text advances while the microphone hears you speaking and pauses when you stop. It knows that you're talking, but not what you're saying.
Voice-following. The software listens to the words themselves, matches them against the script, and keeps a highlight on the word you're actually saying. The scroll has no speed of its own; it has your speed.
Why constant scroll means performing to a metronome
Constant scroll asks you to answer an impossible question before the take starts: how fast will I speak? Not on average — at every moment. You'll naturally speed through the setup of a point and slow into the point itself; a fixed scroll averages both into one tempo, and the read flattens to match.
Then there's the error case. Pause longer than the scroll budgeted, and the text drifts ahead; ad-lib a sentence, and it's gone by the time you come back; stumble, and the script continues without you. On a beam-splitter this is particularly unforgiving, because the entire benefit of the device is unbroken eye contact — there's no natural moment to glance at a control, and the recovery from a runaway scroll is either chasing it or stopping the take.
None of which makes constant scroll useless — more on that at the end — but it explains the wooden quality of many prompted reads: the speaker's attention is split between saying the words and tracking a conveyor belt.
What voice-following looks like on this device
TellyPrompter runs in a browser, so putting it on an Elgato Prompter is just dragging a tab onto that display and going fullscreen. From there the experience changes shape: you read, and the highlight follows the word you're saying. Pause to think, and it waits. Skip a paragraph you've decided against, and it picks you up where you actually resumed. Go back to retake a sentence, and it goes back with you.
The practical difference on this hardware: your delivery stops serving a number. The Elgato's screen is close to your face by design, so a scroll mismatch is very visible; a highlight that's simply always on your word removes the mismatch as a category. Voice commands cover the moments your hands are on a keyboard out of frame or not available at all — saying "Telly, stop" ends a take, "Telly, again" retakes it.
Speech recognition runs on the device and the audio is never uploaded.
Two honest limits, because they're the browser's rather than anyone's marketing. Voice-following depends on the Web Speech API, which Firefox doesn't implement — in Firefox everything else works, and the constant and sound-level modes still scroll, but the listening mode isn't available. And recognition quality belongs to the browser, not the prompter: it varies by platform, language and microphone, and no prompter software can tune it. If a specific language matters to you, the two-minute test below settles it faster than any spec sheet.
Browser-based vs installed, and shoot day without wifi
A reasonable instinct says an installed desktop app must be more dependable than a website. For prompting on this device, the distinction is thinner than it looks.
A browser prompter needs no install, no account and no updates to manage, and it works identically on the laptop in your studio and the borrowed machine at a client's office. The traditional weakness — needing a connection — is solved by the installable web app pattern: TellyPrompter installs as a PWA, and after the first online visit the prompter itself works offline. A shoot location with unreliable wifi is a normal place to open a teleprompter, and the software is built on that assumption.
The genuinely browser-shaped limitation is the one above: the recognition engine is whichever one your browser provides. An installed suite can ship its own.
Honest fit: when a paid desktop suite is the better answer
If your work looks like broadcast — a second operator, hardware scroll controllers, timecode, multi-prompter output to several screens at once — the professional desktop suites are built for exactly that, and a browser tab is not trying to be. Similarly, if you must prompt in Firefox specifically, or you need guaranteed recognition in a language your browser handles poorly, an app that bundles its own recognition may serve you better.
But if you're the person the Elgato Prompter is mostly sold to — one creator, one camera, recording yourself to a lens — the questions that decide your software are simpler. Does the text move at your pace or its own? Does it survive a pause, an ad-lib, a stumble? Does it work in the room with bad wifi?
The test costs two minutes: put a tab on the prompter's screen, paste a real script, and read it once aloud — including one deliberate pause and one deliberate ad-lib. Whichever software keeps your place through both has answered the question. The hardware you've already chosen well; this is the other half. And once the software is settled, most of what remains is delivery — which we've covered in how to read a teleprompter without sounding wooden.