A built-in teleprompter is the upgrade most screen recorders are missing
For six months I rotated through four setups: a teleprompter app on a second monitor, a script taped above the webcam, the script in a Notes window I'd keep mostly off-camera, and reading from memory. None worked. The fifth setup — a teleprompter inside the recorder — did. Here's the difference.
What's wrong with second-screen prompters
Your eyes go off-camera. Even slightly. The webcam is one foot in front of you; the second monitor is at a 30–45° angle. The viewer can tell you're reading. Specifically: they can tell you're not looking at them.
"The webcam is one foot in front of you. The second monitor is at a 30–45° angle. The viewer can tell you're not looking at them."
This is fine in a podcast, where there's no expected eye contact. It is not fine in a talking-head clip designed to feel direct. Direct is the whole point of a talking head.
What's wrong with paper above the webcam
You can't scroll. You either reset your script every 30 seconds or you tape an entire page above the laptop and squint. The squinting is worse than the eye drift.
What's wrong with a window on the same screen
The window shows up in the recording. Either you put it in a corner the recorder doesn't see (good luck cropping consistently), or you accept that your viewer will see a translucent Notes window with your script. Not great.
What "built-in" means and why it works
A built-in teleprompter is a window on the recording machine that is invisible to the recording itself. On macOS, this works via the system's NSWindowSharingType exclusion — the window exists, you can see it, but the screen capture API doesn't. The flag is documented, the behaviour is platform-supported, and many native macOS apps use it for similar reasons.
Three things change about your delivery:
- Your eyes stay on the camera. The prompter sits right under the webcam; the drift is sub-degree.
- You can self-pace. Modern teleprompters scroll on a keystroke, on a foot pedal, or automatically based on your reading speed. You stop racing.
- Your script is in the recorder, not in another app. Take notes between takes. Edit a line in place. Don't alt-tab.
The reading-speed problem
natural speaking speed vs. ~200 wpm reading speed — set your prompter scroll to speaking speed, not reading speed
Most people read written English at ~200 words/minute and speak conversational English at ~130. If you read your script at reading speed, you sound like a news anchor. If you slow down to natural speech, your prompter outpaces you, you lose your place, and the take falls apart.
Two fixes. Either: scroll on a keystroke (most reliable, slightly clunky), or set the scroll speed to your speaking speed (less than reading speed; experiment). For 130 wpm, that's roughly one line every 4 seconds at typical line length.
The script itself
A few things I had to relearn:
- Short sentences. Spoken English doesn't have semicolons.
- Contractions. "Don't" reads as natural; "do not" reads as scripted.
- One idea per line. Line breaks in your script become breath beats in your delivery.
- Mark your emphasis. Bold the word you want stressed; you'll do it without thinking on the take.
The script that reads well silently rarely reads well aloud. Read it aloud once before you record. The lines that taste wrong in your mouth will sound wrong in the take.
Which recorders ship this
Surprisingly few. Loom and Tella don't ship it; Screen.studio doesn't ship it; OBS has community plugins of varying quality. CursorFlow ships a teleprompter that's invisible to the recording, with adjustable speed and a keystroke-step mode.
It's a small feature that changes the output of a talking-head shot more than almost anything else on this blog. Combined with the rest of the production checklist, it's the difference between a video that sounds like you and a video that sounds like you reading.
