Which Microphone Works Best With Voice-Follow Teleprompter Scrolling on iPhone: Built-in Mic vs AirPods vs Lavalier

Most microphone advice for creators is about how the video sounds. With a voice-driven teleprompter there is a second listener: the speech recogniser that moves your script. In Teleprompter: Camera Overlay, the free iPhone app this series is written around, on-device speech recognition matches your spoken words against the script and scrolls at your pace, so what the microphone feeds it decides whether the text glides or stutters. Yesterday's post on teleprompter settings for different lighting ended by promising this one.

One thing this post does not claim: I am not going to describe the app's internal mic-selection logic, because I would be guessing. Everything below is about what you can verify yourself in a minute: how clean the audio reaching the phone is, and what that does to the recognition and the recorded track.

Why the mic changes how the script scrolls

Speech recognition is good at close, clean, consistent audio and gets steadily worse as the signal gets echoey, distant or uneven. A crisp, close voice is matched word by word; the same sentence bouncing off a tiled kitchen arrives smeared, and the recogniser either waits for a word it can trust or guesses. On a teleprompter that shows up as a scroll that lags a beat behind you, or jumps when a guess lands on a later line. The app re-finds your place if you stumble or jump back to the top, which covers the occasional miss, but a mic that loses every third word turns every sentence into a stumble.

The second effect is that mic placement changes the video's audio track in the same direction. Close and consistent sounds professional; distant and echoey sounds like a phone in a room. A mic choice that helps the scroll usually helps the final clip too. Where it becomes a trade-off is the Bluetooth and convenience column below.

Built-in mic vs AirPods vs wired lav vs wireless lav

Four options, judged on cost, convenience, what they do to recognition and how they sound in the saved video. General tendencies, not measurements; your room will move them.

• Built-in iPhone mic. Cost: nothing. Convenience: nothing to charge, pair or clip. Recognition: good at arm's length in a quiet room; drops off as you move back, and picks up room echo, the fridge and the street. Sounds: clean but roomy; obviously a phone in a bare room, fine in a soft furnished one.
• AirPods (or similar earbuds). Cost: you probably own them. Convenience: high, and you can stand further from the phone. Recognition: the mic sits near your mouth, so the voice is close and consistent, which is what the recogniser wants. Sounds: Bluetooth audio used for video recording tends to be compressed and lower quality than a wired mic, so the track can sound thin, and the earbuds are visible on camera.
• Wired lavalier. Cost: low to moderate, plus the right adapter for your iPhone's port. Convenience: a cable to hide, but no battery and no pairing. Recognition: close, constant level, the best case for voice-follow. Sounds: the classic clean talking-head track most tutorial videos use.
• Wireless lavalier. Cost: the highest of the four. Convenience: free movement, but two batteries and a pairing step. Recognition: the same close, consistent voice as a wired lav. Sounds: as good as wired on a good system; the risks are dropouts and the receiver's connection.

Placement matters more than brand. A lav clipped 15 to 20 cm below the chin, pointing up, gives a consistent level whether you turn your head or lean in. Clip it on a collar edge and you get fabric rustle every time you move. Outdoors, a foam or furry windshield is not optional: wind on a bare lav sounds like a storm to a human and like silence with random words to a recogniser. Handling noise is the indoor equivalent; if you hold the phone, the built-in mic hears your fingers, so use a stand.

A 60-second test to pick a mic for your room

• Load a script you know, put the phone where it will record, and connect the mic you want to test.
• Read the first paragraph at your normal speed, looking at the lens, and watch the scroll. Does it stay a line ahead, or trail you?
• Say one sentence, stop for three seconds, then skip a line and carry on. The text should wait, then find your new place.
• Record 15 seconds, play it back with earphones, and listen for echo, hiss, rustle and traffic. That track is what your viewers get.
• Repeat with the next mic. Keep the one where both the scroll and the playback were calm. In a small, soft room that is often the built-in mic; in a bigger or harder room it is usually a lav.
• Write the winner down for that room, next to the brightness and text size you noted in the lighting post.

When the scroll lags or jumps: fixes before you buy anything

Most scrolling complaints are not the microphone. Try these first, in order:

• Say the words as written. Voice-follow matches your speech to the script; if you paraphrase, it has nothing to match. Edit the script to how you actually talk.
• Pause between sentences. A short gap gives the recogniser a clean boundary and gives you a natural edit point.
• Get closer, or get the mic closer. Halving the distance is the cheapest upgrade there is.
• Kill the background. Fan, fridge, open window, music. A recogniser hears all of it.
• Check the language. The app detects the script's language and listens in it; a stray paragraph in another language will confuse that.
• Switch to auto-scroll for the take. One tap puts the script on a constant speed you set with the slider; tap to pause, drag to jump. Fix a bad line in the takes library later rather than fighting the scroll mid-sentence.

When NOT to use voice-follow

Voice-driven scrolling is the right default for one person in a quiet room. It is the wrong tool in three situations, and the answer in each is auto-scroll with the speed slider, which the voice-follow vs auto-scroll post compares in detail.

• A noisy café or street. Even a lav picks up the surroundings, and the recogniser has to separate you from the espresso machine. Auto-scroll does not listen, so it does not care.
• More than one speaker. Two voices and one script means the recogniser can match the wrong person's words. Set a speed, rehearse once, and let it run.
• Scripts full of names, brand words or technical terms. A product name the recogniser has never heard gets transcribed as something else, and your scroll waits for a word that never arrives. Spell those words phonetically in the script, or switch to auto-scroll for that section.

The rest of the kit does not change with the mic. The app records portrait 4K from the front or back camera while the script floats over the preview next to the lens, the script never appears in the saved video, and recognition is on-device: scripts and recordings never leave the phone, and it works in airplane mode. It is iOS-only, free with an occasional ad, and a small one-time purchase removes ads forever; no subscription, no watermark. For the camera and light side of the same setup, the talking head video setup guide covers stands, distance and framing.

Next in the series

Everything so far assumes you are the one on camera. The next post flips it: how to use a teleprompter when you are NOT the one being filmed, for interviews and two-person videos.