YALTI Prompter

Your words. Your voice.

A minimalist, open-source teleprompter for Windows that flows out of the top of your screen and follows your voice — entirely offline.

---

YALTI Prompter sits at the top center of your screen, right under your webcam, as a soft dark surface that seems to grow out of the bezel — inspired by the Dynamic Island. Load a script, start speaking, and YALTI listens on your own computer and keeps the line you are reading in view.

It understands that people don't read perfectly. Pause, skip a sentence, repeat yourself, ad-lib, or wander off-script for a story: YALTI waits, and when you come back to your words — even at a completely different part of the script — it finds your place again.

Features

• Follows your voice. Local, open-source speech recognition (sherpa-onnx) tracks where you are in the script. It handles skipped sentences, repeats, paraphrasing, long pauses, going off-script and resuming elsewhere, and only moves when it is confident. If it ever jumps somewhere you didn't mean, one click (or Backspace) takes you back.
• A liquid island, not a window. The prompter emerges from the top edge, expands, folds into a compact pill and retracts into the bezel with one continuous, spring-driven motion. Choose Liquid, Flush or Floating styles and tune the intensity.
• Readable everywhere. Nine bundled open-source fonts, adjustable size, weight, line and letter spacing, alignment, colors, highlight color, background opacity and reading-line position. Mirror mode for beam-splitter glass.
• Scripts the way you write them. Open .txt, .md (rendered cleanly — no Markdown symbols), .srt and .vtt files; paste from the clipboard; drop a file — or text highlighted in any app — on the island or in Settings; or write in the built-in editor. Files reload automatically when you save them elsewhere, and YALTI remembers where you were in each script. Text in [square brackets] becomes a quiet stage direction that is never tracked.
• Auto-scroll and manual control whenever you prefer: adjustable speed, optional countdown, mouse wheel, keys, or drag the text.
• Out of your way. Transparent areas click through to the apps beneath. Always-on-top (even over full-screen slides), system-tray access, and global shortcuts that work while another app has focus.
• Private by design. No accounts, no cloud, no telemetry, no recordings. Scripts and audio never leave your computer, and it works fully offline. The only time it goes online is to check GitHub for updates — when you ask it to, or if you turn on automatic updates.
• Keeps itself up to date — if you want. Settings → Updates checks GitHub for a new version, downloads it, verifies it against the release's published checksums and installs it. Automatic updates install when you quit, never in the middle of a talk.
• Open source throughout — code (MIT), speech engine and model (Apache-2.0), fonts (OFL), icons (ISC). See THIRD_PARTY_NOTICES.md.

A closer look

Compact — fold the prompter into a small pill with status and progress.	Notifications — the island briefly grows to tell you something.
Quiet controls appear only when you hover.	Make it yours — fonts, colors, size and the top-edge style.
Settings apply live while you adjust them.	Built-in editor updates the prompter as you type.

Download and install

Get the latest version from the Releases page. No account or developer tools are needed.

File	What it is
YALTI-Prompter-Setup-x.y.z.exe	Recommended. Installer for your user account (no administrator rights needed). Adds Start menu and desktop shortcuts and an uninstaller.
YALTI-Prompter-Portable-x.y.z.exe	Single-file portable app. Run it from anywhere (e.g. a USB stick); settings are stored next to it.
YALTI-Prompter-x.y.z-win-x64.zip	Portable folder. Unzip and run YALTI Prompter.exe. Starts faster than the single-file portable.

Requirements: Windows 10 or 11 (64-bit) and a microphone for voice tracking.

Windows SmartScreen: the release builds are not code-signed yet, so Windows may show “Windows protected your PC”. Choose More info → Run anyway. You can verify a download against the SHA-256 checksums published with each release.

Update from Settings → Updates (or Check for updates… in the tray menu), or turn on automatic updates there. Running a newer installer over an existing installation also works: it notices YALTI is already installed and offers to update it, to check for updates, or to uninstall.

Uninstall from Settings → Apps → Installed apps → YALTI Prompter, or with the uninstaller in the Start menu. Your settings stay in %APPDATA%\YALTI Prompter unless you delete that folder.

Quick start

1. Launch YALTI Prompter. The island flows out of the top of your screen with a short welcome script.
2. Open your script with Ctrl+O, paste one with Ctrl+V, or drop a .txt / .md file — or highlighted text — on the island.
3. Press Space (or the microphone button) and start reading aloud. YALTI follows along.

Hover over the island to reveal the controls. Press ? for every shortcut.

Everyday controls

Action	In the prompter	From anywhere (global)
Start / pause (listen or auto-scroll)	Space	Ctrl+Alt+S
Voice tracking on / off	M	Ctrl+Alt+M
Show / hide the prompter	H	Ctrl+Alt+Y
Expand / collapse	Enter / Esc	Ctrl+Alt+Enter
Scroll a line / a page	↑ ↓ / PgUp PgDn	Ctrl+Alt+PgUp / Ctrl+Alt+PgDn
Slower / faster auto-scroll	[ / ]	Ctrl+Alt+- / Ctrl+Alt+=
Back to the start / end	Home / End	Ctrl+Alt+Home
Jump back after voice tracking jumps	Backspace	Ctrl+Alt+Backspace
Text size	Ctrl + / Ctrl −
Open · Paste · Edit · Reload script	Ctrl+O · Ctrl+V · Ctrl+E · Ctrl+R
Settings	Ctrl+,

Global shortcuts can be changed or turned off in Settings → Shortcuts. Drag the island sideways to move it along the top edge; drag either bottom corner to resize it.

The full guide is in docs/USER_GUIDE.md.

How voice tracking works

YALTI keeps the last few recognized words and aligns them against the whole script with a fuzzy, phonetics-aware alignment that tolerates misheard, skipped, extra, merged and split words. Nearby continuation is cheap; moving far away needs strong, unambiguous evidence that repeats over several words; and when recent speech matches nothing, YALTI holds still and quietly shows that it is listening for your script. Details and tuning notes are in docs/ARCHITECTURE.md.

Everything runs locally: a 70 MB streaming speech model (English) uses about 5–10 % of one CPU core while you speak and is unloaded a few minutes after you stop.

Building from source

git clone https://github.com/silent-diffusion/YALTI-Prompter.git
cd YALTI-Prompter
npm ci
npm run fetch-model   # downloads and verifies the speech model (~73 MB)
npm start

Create the installer, portable app and zip in dist/ with npm run dist. Prerequisites, tests and release steps are in docs/BUILDING.md and docs/RELEASING.md.

Technology

• Electron with plain HTML, CSS and JavaScript modules — no UI framework
• sherpa-onnx streaming Zipformer transducer for on-device recognition, running in an isolated utility process
• marked for parsing Markdown scripts
• Fontsource variable fonts and Lucide icons
• electron-builder + NSIS for the installer

Every runtime dependency and its license is listed in THIRD_PARTY_NOTICES.md, and npm run check-licenses verifies them against an open-source allowlist.

Contributing

Bug reports, ideas and pull requests are welcome — see CONTRIBUTING.md and our Code of Conduct. Please report security issues privately as described in SECURITY.md.

License

YALTI Prompter is released under the MIT License. Bundled third-party components keep their own open-source licenses; see THIRD_PARTY_NOTICES.md.