If your teleprompter lags behind what you’re saying, the delay is almost always in the speech recognition engine, not the scroll itself. Cloud-based voice recognition — which most browser-based teleprompters use — adds 200 to 500 milliseconds of latency because your audio has to round-trip to a server and back. On-device speech recognition runs at 50 to 100 milliseconds, two to ten times faster. This post explains the technical reason behind teleprompter lag, the two real fixes, and why focus-line position matters more than most people realise.

Why speech recognition is the source of the lag

When you speak into a voice-tracked teleprompter, a pipeline runs behind the scenes. Your microphone captures audio. That audio gets sent to a speech recognition engine. The engine returns text. The app matches the returned text against your script and updates the scroll position. Each stage adds time.

The stage that adds the most time by far is speech recognition itself — and the specific bottleneck is whether that recognition runs on your device or on a server somewhere else.

Cloud vs on-device speech recognition

Cloud-based speech recognition — used by browser teleprompters via the Web Speech API, or by apps calling services like Google Cloud Speech-to-Text — sends your audio over the internet to a server, waits for the server to process it, and sends the text back. Even on fast broadband, this adds 200 to 500 milliseconds of latency per word. On poor connections, it can be worse. When your connection drops, it stops working entirely.

On-device speech recognition — used by apps built on Apple’s SpeechAnalyzer (iOS 26+), Apple’s older SFSpeechRecognizer, VOSK, or similar engines — processes audio locally on your device. Modern phone and tablet processors can run recognition in 50 to 100 milliseconds. Two to ten times faster than cloud, and it works offline.

The difference shows up as perceived lag. At 300ms of delay, you end up speaking roughly one word ahead of the prompter, which is just enough to throw off your reading. At 50ms, you’re effectively locked in sync.

Fix 1 — switch to on-device voice tracking

This is the real fix. An app with native on-device voice tracking will have noticeably lower lag than any browser or cloud-based solution. You can test which category an app falls into quickly: put your device in airplane mode and try voice-activated scrolling. If it keeps working, it’s on-device. If it stops, it’s cloud-based.

Steady Cue runs voice tracking natively — scroll keeps pace even offline, with latency in the 50–100ms range.

Fix 2 — adjust your focus line position

The second fix is simpler and independent of the engine: move your focus line down the screen.

Your focus line is the position where you’re actually reading from — the line of text your eye is tracking. On most teleprompter apps this is adjustable, though many users leave it at the default. The default is often at the top of the visible text area, which means the script has nowhere to scroll to before you catch up to the last word. Any engine lag shows immediately.

Move the focus line to around 40% of the screen height. Now there’s a buffer of roughly half a screen of upcoming text below the line you’re reading. Even with some engine lag, the words stay ahead of your voice and the read feels natural.

This is a setting-level fix that works on any app, with any engine. Even if you can’t switch teleprompters, adjusting the focus line alone usually resolves 80% of perceived lag.

What 40% actually looks like

Imagine your screen divided into five horizontal bands. Your focus line should sit at roughly the second band from the top — so about 40% down from the top edge. Above your focus line is the text you’ve already read (often greyed out). Below it is text scrolling up toward you.

If you’re using a beam-splitter rig, this position also happens to be where your eye line is most natural for looking into the camera lens.

The bottom line

Teleprompter lag is almost always engine latency, not the scroll. Fix the engine — switch to on-device voice tracking. Fix the framing — move your focus line to around 40% of the screen. Do both and the lag disappears.