Most teleprompters are built for English and retrofitted for everything else. TellyPrompter treats 34 languages, right-to-left scripts and dyslexia-friendly reading as first-class from the start: paste a script in Arabic, Hebrew, Hindi, Japanese or Welsh and the app detects the language, sets the reading direction, and — where the browser supports it — follows your voice as you read, all on the free tier.
This piece covers what "supports your language" should actually mean in a teleprompter, why right-to-left is harder than it looks (especially on a mirror rig), where voice-following stands across languages honestly, and why the typeface question belongs in the same conversation.
What language support should mean
A prompter's language support is often a checkbox that means "the text renders". For a tool you perform from, the bar is higher, because four separate things have to be true:
The script displays correctly — including non-Latin alphabets, connected scripts like Arabic where letters change shape by position, and the diacritics many languages depend on.
The direction is right — not just per-line, but as a reading experience: where new text arrives from, which edge the eye returns to.
The reading aids work in your language — sentence-end markers, the position highlight, the line the eye rests on. A highlight that only understands full stops fails everywhere punctuation differs.
And ideally, the following works — the voice-tracking that keeps your place, in your language, not just in English.
TellyPrompter handles 34 languages across that whole stack. When you paste a script, automatic detection identifies the language and sets sensible defaults — direction included — and an explicit language choice always wins over the detection, for the cases where you know better than the statistics.
Right-to-left, done properly
Arabic, Hebrew, Persian, Urdu: hundreds of millions of people read right-to-left, and RTL is where retrofitted prompters show their seams. Text alignment is the easy part. The harder parts:
Layout mirrors as a whole. In an RTL reading experience, the eye's home edge is the right, progress runs leftward, and the interface around the script should agree with the text rather than fight it. TellyPrompter has full RTL support across those 34 languages, non-Latin scripts included — the direction is a first-class property of the script surface, not a text-align property bolted on.
Mixed-direction text is normal, not an edge case. Real scripts mix directions constantly — an Arabic script quoting an English product name, a Hebrew script with a number and a URL. This is exactly why automatic detection alone isn't enough and the app gives you an explicit direction control: for genuinely mixed material, you decide the base direction, and the embedded runs sit inside it.
And then the mirror rig flips everything again. Here's the compound case retrofits never survive: a beam-splitter reflects the whole screen horizontally, so an RTL script on prompter glass is a right-to-left text inside a left-right-flipped display. Because mirroring and direction are both first-class in TellyPrompter, the combination just works — mirror mode flips the surface, the script keeps its correct reading direction in the glass, and nothing in the app depends on left or right meaning anything, precisely because mirrored rigs are normal use.
Voice-following across languages, honestly
The voice-follow — the highlight that tracks the word you're speaking, through pauses, ad-libs and backtracking — works in any of the 34 languages the browser's speech recognition supports. Read in Spanish, the script follows in Spanish; the same in Arabic, Korean or Polish.
The honest part, stated plainly: recognition quality is the browser's, not ours, and it varies by language, platform and microphone. The engines behind the Web Speech API are trained more heavily on some languages than others; a language the engine handles well gives you following as solid as English, a language it handles poorly gives you looser tracking. This cannot be tuned from the app's side, and any prompter claiming otherwise about browser-based recognition is overclaiming.
Three things keep this from being a gamble:
The test is two minutes. Paste a real script in your language and read a paragraph aloud. The highlight either holds or it doesn't — you'll know before the first take, not during it.
Auto voice settings resolve sensibly. Untouched, the voice preference switches on for English and stays off elsewhere with a plain explanation; your explicit on/off choice wins in every language, so a well-supported language is one toggle away.
And the other two modes are language-blind. Constant-speed scrolling and sound-level scrolling work identically in every language, every script, every direction — they don't interpret words at all. Whatever recognition thinks of your language, you have a working prompter; the product degrades, it doesn't block. One browser-level note: Firefox doesn't implement the Web Speech API in any language, so voice-following isn't available there — the other modes are.
Where the following does hold, everything built on it comes along: pauses free, ad-libs free, and voice control of the take itself.
The typeface is part of language support
Reading aloud under pressure is its own skill, and for a lot of people — dyslexic readers prominently among them — the typeface decides whether a script is performable at all.
TellyPrompter ships OpenDyslexic as a first-class script face: same size range, same reading states, same treatment as the default face, not an accessibility toggle tucked in a corner. For readers who find standard faces swim under pressure, that's simply the difference between a script they can perform from and one they can't.
Two related choices run through the whole app, and they matter more on a rig than anywhere else. No cue is carried by colour alone — roughly one man in twelve can't reliably resolve a colour-only distinction, and the position highlight is the single most load-bearing element on the screen, so it never depends on colour to be findable. And contrast is set assuming beam-splitter glass has already eaten a third to a half of the light, so the critical cues survive the rig, whatever script they're written in.
Setting up in your language: the two-minute pass
- Paste your script. Detection sets the language and direction; override either explicitly if the material is mixed or the detection guessed wrong.
- Pick the face — the default or OpenDyslexic — and size it from where you'll actually stand, not from the desk.
- On a mirror rig: mirror mode on, then confirm with a real sentence in the glass that the script reads correctly in its own direction.
- Read one paragraph aloud. If the highlight holds, voice-follow is available in your language; if it's loose, switch to sound-level or constant and the session proceeds identically.
- Take the take.
The larger point is the quiet one. A creator recording in Arabic or Hindi or Hebrew shouldn't need an English-first tool's grudging accommodation — the reading experience, the rig behaviour and the honesty about recognition should all be built as if their language were the first one considered. That's a page most of the category can't write truthfully, which is exactly why it was worth building. For where the prompter fits in the wider recording setup, start with the YouTube setup guide; for the craft of the read itself, how to read a teleprompter without sounding wooden applies in every language here.