A teleprompter for online language teachers — one lesson, two languages

Teaching a language on camera means speaking in a way no prompter expects. Your model sentences are deliberately slow. You stretch a vowel so learners can hear where it sits, say a phrase twice at half pace, and leave a silence where the student is supposed to answer. A prompter scrolling at a fixed speed treats all of that as falling behind — by the second repetition the text is a paragraph ahead of your mouth, and the workaround, speeding up to catch it, is the one thing a diction model must never do.

Then there is the two-language problem. The same lesson often exists twice: once with explanations in English for beginners, once in the target language for immersion learners. That is two scripts recorded in two sessions, and the recognition that followed your English in the morning has to follow your Spanish in the afternoon. Most tools quietly assume you speak one language at a prompter.

If you teach Hebrew or Arabic, add a third problem: most prompter apps were built for left-to-right text. Paste a Hebrew script in and the punctuation jumps to the wrong end of the line, the alignment fights the reading direction, and the tool you use on camera undermines the writing system you teach.

And all of it lands on one person. There is no crew: you are the presenter, the camera operator and the editor, recording between lesson slots that students booked at fixed times. A lesson that takes an evening of editing does not get recorded.

The usual answers each cost something. A fixed-speed scroll means rehearsing until your teaching pace matches the machine — the tail wagging the dog, with slow, clean diction as the first casualty. A foot pedal or a partner scrolling by hand fixes the pace and occupies a limb, or a second person, for every recording.

Some teachers pre-record the model sentences separately and edit them into the lesson afterwards. It works, doubles the editing on every video, and the join is audible where the room tone changes.

For the two-language versions, most keep two documents in two places, email scripts to themselves between the tablet on the desk and the phone on the tripod, and rely on file names like lesson-4-spanish-FINAL to keep the pair together.

And for right-to-left scripts the common answer is not using a prompter at all: printed pages beside the camera, or memorised bullet points — in exactly the lessons where a precise model sentence matters most.

Voice follow reverses the relationship: the script follows you, not a clock. Speak a model sentence at half pace and the text moves at half pace. Stop entirely — a silence for learners to answer into — and it waits. Say the phrase again and it does not run off without you, and when you skip the aside you meant to cut anyway, it catches up. Slow, exaggerated diction is precisely the delivery it tolerates best, because there is no speed to fall behind.

The recognition runs on the device, in the language the device recognises speech in — so the Spanish version of the lesson is followed as Spanish and the English version as English. It is worth checking once, before planning a course, that your phone recognises every language on your list. Nothing you say is uploaded to make any of it work.

Hebrew and Arabic scripts simply lay out right to left, with no setting to find. Alignment, margins and punctuation land where a reader of the language expects them. The rest of the typography earns its keep in the same lessons: text size, line spacing, side margins and five faces including Lexend and OpenDyslexic, because a script read mid-lesson, at arm’s length, wants to be larger and looser than prose.

Not every lesson is continuous prose. For drills and vocabulary runs there is manual and timed scrolling: drive the script yourself, or set a pace and let it run, with a progress bar showing the time remaining — useful when the recording has to fit the slot your platform sells.

Scripts live in one library, in folders — one per course, or one per language — each with a word count and an estimated read time, so the immersion version and the explanation version sit side by side instead of in two apps. With cloud sync, part of a subscription, the pair you wrote on the tablet at your desk is already on the phone on the tripod, in the same wording, with no file emailed to yourself.

A teacher sells a beginners’ Hebrew course and records at home between live lessons. Lesson 12 exists twice: an English-explanation version and an immersion version, two scripts side by side in the course folder. She records the English version first. Then she opens the Hebrew script — already laid out right to left, without touching a setting — checks that the device recognises Hebrew, and records the immersion version in the same sitting.

The model sentences come out the way she teaches them: slow, stretched, said twice. The prompter waits through every repetition and every answer-length silence instead of racing ahead, so she never once speeds up to chase the text — the habit that would undo the lesson. Her students hear clean diction at learner pace, and the recording took one sitting, not an evening of editing. Next week, planning on the tablet, both scripts are already there, synced, ready for lesson 13.