The teleprompter

What the teleprompter does

The teleprompter is Rhetoric's rehearsal stage. You load one of your speeches, speak it aloud, and the text moves with your voice. It waits when you pause and advances when you continue. When you go off script, a recovery panel helps you find your way back. When you finish, you get a report of what the session measured.

Rhetoric never stores your audio. The voice signal is processed transiently for transcription, used to advance the script, and discarded. What is saved after a session is the metrics: duration, pace, coverage, and departures.

The prep screen

When you open the teleprompter from your Library or the editor, the prep screen shows your speech title, word count, estimated duration, and section count. Below that, readiness rows tell you exactly what the session will use.

1. 1The tracking row says which engine will follow your voice. "Enhanced tracking" means a managed speech engine with smoother word timing. "Voice tracking ready" means your browser's own recognizer. If voice is unavailable, the row says so and the session uses timed scroll instead.
2. 2The recovery cards row says whether your cards were written by the analysis engine ("AI recovery cards ready") or built locally from your script ("Draft recovery cards (local)"). AI cards generate automatically when you save in the editor.
3. 3The waypoints row lists your speech sections. During the live session these become tappable jump targets with live time estimates.

Every label on the prep screen names what is real: which engine, which card source, and that audio is never stored.

Engines: enhanced and browser

Paid sessions use the enhanced tracker where possible. It sends your microphone audio to a speech engine over a secure connection, returns word-level transcripts, and closes when the session ends. If the enhanced engine cannot start, the session falls back to the browser engine and says so in the tracking pill.

The browser engine uses your browser's own speech recognition. It works without any external connection and your audio never leaves your device. It is also the engine behind the free public demo, which is a lightweight sample of voice tracking, not the full paid experience.

Both engines report which one is active. There is no silent switch.

Recovery cards

Every sentence of your script has one recovery card, generated before the session starts. A card has three parts: a lead-in Anchor (a short phrase you can say aloud to step back into the script), a Bridge Question (a rhetorical question that buys you time), and a Thematic Cue (keywords of the sentence ahead).

• AI cards are written by the analysis engine when you save your speech in the editor. They are labelled "AI recovery cards ready" on the prep screen and "AI cards" on the live panel.
• Local draft cards are built from the text of your script without any AI call. They are labelled "Draft recovery cards (local)" on the prep screen and "local drafts" on the live panel.
• The source badge is always visible during the session so you know which kind you are seeing.

Waypoints and settings

Waypoints are the sections of your speech shown on the prep screen and in the side panel during the session. On a desktop they appear as a column on the right. On smaller screens they sit in a bar at the bottom. Each waypoint shows its label and, once the session is live, a time estimate: done, now, or a countdown. Tap any waypoint to jump to that section.

Stage settings open from the gear icon in the top bar during the session. They control text size (22 to 64 px), text tone (Cool white or Warm white), Mirror text for a physical beam-splitter glass teleprompter, tracking mode (Voice or Timed scroll), and timed speed (90 to 180 words per minute).

Settings are saved to this device, not to your account, so each device keeps its own preferences.

The live stage

Click Begin teleprompter to start. The screen goes dark and your script fills the view in large type. A thin reading line sits in the upper third of the screen, underlining the line you are speaking now. Text ahead of you is bright below it, and the lines you have already spoken fade above it.

The top bar shows elapsed time, your current pace in words per minute, estimated time remaining, and a tracking pill. The pill reads "tracking your voice" when the engine is listening, "timed scroll" in timed mode, or a warning when the mic is blocked or unavailable.

In voice mode the text advances when the engine matches your spoken words. It waits when you pause. Silence is not a fault. In timed mode the text glides at a steady rate you set in the stage settings. You can pause and resume timed scroll from the top bar.

Going off script

When you improvise or lose your place, the tracker detects the departure and an amber badge appears at the top of the stage reading "off script, holding your place". The text stays where you left it: the bright words show exactly where to pick up.

On a desktop the side panel swaps from waypoints to the recovery card for the next sentence ahead, headed "Find your way back". On smaller screens the recovery card slides in beside the waypoints so you can see both. Speak a few words of your script and the stage picks you back up with no tap needed.

Manual scroll

Scrolling by hand is navigation, not an interruption. While you scroll, the text follows your finger or wheel. When the scroll settles, the tracker re-anchors at the first fully visible sentence and the reading line snaps there. There is no resume button. If you were off script, manual scroll also clears the departure state.

Presenting in a video call

On a laptop or desktop, voice tracking and a video call can share the microphone. Run both normally. On phones and tablets, only one app can use the microphone at a time. Starting voice tracking on those devices takes the mic away from your call.

The prep screen warns you when this applies and offers a Switch to timed scroll button. Timed scroll never opens the mic, so your call keeps working. If you have two devices, run the call on one and the teleprompter on the other.

The session report

When you click End session (or Finish once you reach the end), a report shows four numbers.

• Duration is clock time from Begin teleprompter to End session, measured locally.
• Pace is words per minute, measured from the words your voice produced divided by the session time. Timed sessions show "No data" because no voice measurement was taken.
• Coverage is how far you reached in the script as a percentage of total words, measured locally.
• Departures is the count of off-script moments detected by the tracker. Each one lists approximately where it happened and how quickly you recovered. Timed sessions show "No data".

"No data" in a timed session is honest: Rhetoric does not invent a score when no speech was captured.

If the session used voice tracking and you went off script, the report lists each departure, approximately which word it occurred at, and the recovery time, with average recovery time displayed below. Voice sessions also show a pauses count, split into rhetorical pauses (on a sentence end or an authored cue) and mid-sentence lapses.

Sessions save automatically when the report appears. The chip at the top right reads "saved to History" and links to your History view, where you can compare runs over time. Click Run it again to reset the stage for another attempt. Click Done to return to your Library.

Timed scroll sessions

Timed scroll runs the script at a constant speed with no microphone. You choose the pace, between 90 and 180 words per minute, in the stage settings. Coverage still tracks how far the scroll reached. Pace and Departures show "No data" because there is no voice signal to measure.

This is not a limitation to apologize for: timed scroll keeps you rehearsing on any browser, in any video call, and on any device.

When something goes wrong

If voice tracking lags, stalls, or refuses to move, "Troubleshoot teleprompter tracking and pace" walks through the most common fixes: confirming the mic, closing competing tabs, speaking script words to re-anchor, and switching to timed mode so the rehearsal still happens.

If the browser blocks the mic entirely, "Allow or restore microphone access" covers how to fix it in Chrome, Edge, Safari, and Firefox. "Supported browsers, engines, and fallbacks" explains which engines work where. "Use the teleprompter in a video call" covers mic exclusivity on phones and the two-device setup.