Free, open-source, 100% offline AI text-to-speech.

A native desktop app for macOS, Windows, and Linux that reads text aloud with natural-sounding voices.
Everything runs locally on your machine. No cloud, no accounts, no telemetry.

• What it does
• Screenshots
• Install
  • macOS first launch
  • Windows: SmartScreen warning
• Supported languages
• Speaking tips: pauses & shortcuts
• How it works
• API
• Repository layout
• Build from source
• Scripts
• Documentation
• Contributing
• Support
• License
• Credits

• Reads text aloud with 50+ natural voices across 8 languages
• Speech runs 100% offline — your text never leaves your computer
• Built on Kokoro-82M, an open-weight TTS model ranked highly on the TTS Arena
• Accessible Talker mode (type-and-speak) and inline pause tags — see Speaking tips
• Desktop app with menu-bar / system-tray integration
• Tells you in-app when a new version is available (the only network call — checks GitHub releases)
• Chrome & Safari extensions to read any webpage in one click
• Local HTTP API on port 51730 for extensions and scripts

Desktop app. Paste, pick a voice, hit play.

System tray. Menu-bar access while you work.

Grab the latest release for your platform from the Releases page:

• macOS: Out Loud-<version>.dmg (Apple Silicon & Intel)
• Windows: Out Loud-<version>.exe
• Linux: Out Loud-<version>.AppImage

Or build it yourself. See Build from source.

1.0.3 and newer: the app is Developer-ID signed and Apple-notarized. Double-click — it opens. No dialog.

1.0.2 only: that one release shipped signed but un-notarized (Apple's notary service was stuck during the release window). On first launch you'll see "macOS cannot verify the developer of Out Loud". Upgrade to 1.0.3+ to skip this entirely, or work around it once:

• macOS 15 (Sequoia)+: click Done → open System Settings → Privacy & Security → scroll to the Security section → click Open Anyway next to the "Out Loud was blocked" line → authenticate.
• macOS 14 (Sonoma) and older: right-click Out Loud in /Applications → Open → click Open in the dialog.

Either way macOS remembers your decision — future launches are direct.

(If you see "Out Loud.app is damaged and can't be opened" instead, the download likely got corrupted — re-download. Last-resort workaround: xattr -dr com.apple.quarantine "/Applications/Out Loud.app".)

Windows may show "Windows protected your PC" because the installer isn't signed with a paid Windows code-signing cert. Click More info → Run anyway to proceed.

English (US & UK), Japanese, Chinese (Mandarin), Spanish, Brazilian Portuguese, Italian, and Hindi. 50+ voices total.

Pauses. Insert a silence anywhere in the text with a tag — all of these are equivalent and produce a 1-second pause:

[1s]        [1000ms]        <pause=1s>        <break time="1s"/>

Use decimals or ms for finer control ([0.5s], [250ms]). Punctuation also pauses automatically, so normal prose already sounds natural: period/semicolon ≈ 0.4s, colon ≈ 0.3s, comma ≈ 0.2s, and a line break ≈ 0.4s.

Keyboard.

• ⌘/Ctrl + Enter — speak the current text (works any time).
• Shift + Enter — insert a line break without speaking.
• Esc — jump the cursor back to the text box.
• Talker mode (Settings): plain Enter speaks and then clears the box, and the text box stays editable while audio plays — type-and-go, no need to find a button.

Blend voices (API). The voice field accepts mixing formulas, e.g. af_heart*0.6+am_michael*0.4 (weights must sum to 1) — see API.

The Electron app owns the ONNX model, runs TTS in a worker thread, and exposes a local HTTP API on port 51730 that the browser extensions talk to. Nothing goes to the network.

flowchart LR
  subgraph Desktop["Out Loud desktop app (Electron)"]
    UI[Renderer UI<br/>React + Vite]
    Main[Main process]
    Worker[TTS worker<br/>ONNX + espeak-ng]
    UI <-- IPC --> Main
    Main <-- messages --> Worker
  end

  HTTP[["HTTP :51730"]]
  Main --- HTTP

  Chrome[Chrome extension]
  Safari[Safari extension]
  Scripts[Your scripts / curl]

  Chrome --> HTTP
  Safari --> HTTP
  Scripts --> HTTP

The Electron app exposes a local HTTP API on 127.0.0.1:51730. A few quick examples:

# List voices
curl http://127.0.0.1:51730/api/v1/audio/voices

# Generate audio to a file
curl -X POST http://127.0.0.1:51730/api/v1/audio/speech \
  -H "Content-Type: application/json" \
  -d '{"voice":"af_heart","input":"Hello, world."}' \
  --output hello.wav

Full reference: docs/app/api.md. Machine-readable OpenAPI 3.1 spec: docs/app/openapi.yaml. Also served live at http://127.0.0.1:51730/api/v1/openapi.yaml while the app is running.

.
├── electron/           TTS engine + main process (TypeScript)
├── electron-ui/        Renderer UI (React + Vite + Tailwind)
├── chrome-extension/   Chrome / Chromium extension
├── safari-extension/   Safari extension (generated from chrome-extension)
├── tray-app/           Standalone menu-bar app variant
├── scripts/            Icon generation, dev server helpers
├── build-resources/    Icons, entitlements, DMG background for electron-builder
└── docs/               Deeper docs (app, extensions, build)

• Node.js >= 22 (LTS)
• npm >= 10
• macOS, Windows, or Linux

git clone https://github.com/light-cloud-com/out-loud.git
cd out-loud
npm install
npm run electron-ui:install

# Electron app + hot-reloaded UI
npm run electron:dev

# Compile only the Electron main process
npm run electron:compile

# Run the UI in a browser (no Electron) for faster iteration
npm run electron:browser

npm run electron:build           # current platform
npm run electron:build:mac       # macOS .dmg
npm run electron:build:win       # Windows .exe
npm run electron:build:linux     # Linux .AppImage

Output lands in releases/<platform>/. See Output layout.

npm run extension:chrome:pack    # zip the Chrome extension
npm run extension:safari:convert # convert to Safari (macOS, Xcode required)
npm run extension:test           # smoke-test the local HTTP API

All build artifacts land under releases/ (gitignored):

releases/
├── macos/           *.dmg, *-mac.zip
├── windows/         *-Setup.exe
├── linux/           *.AppImage, *.deb
└── extensions/
    ├── out-loud-chrome.zip
    └── safari/      Xcode project (generated)

For automated releases via GitHub Actions, see docs/build/releasing.md.

Deeper docs live in docs/:

• Release notes: CHANGELOG.md — per-version fixes
• App: docs/app/architecture.md, docs/app/api.md, docs/app/voices.md, docs/app/openapi.yaml
• Extensions: docs/extensions/testing.md, chrome-extension/README.md, safari-extension/README.md
• Build: docs/build/releasing.md, docs/build/mac-app-store.md
• Contributing: CONTRIBUTING.md

Issues, pull requests, translations, and new voices are all welcome. See CONTRIBUTING.md to get started.

Out Loud is free and open source. If it's useful to you, you can support its development:

buymeacoffee.com/julia_hk

MIT. See LICENSE.

Bundled third-party components (Kokoro-82M, espeak-ng, onnxruntime-node, Electron, ffmpeg, fonts) retain their own licenses. See THIRD_PARTY_NOTICES.md for the full list.

• Kokoro-82M, Apache 2.0: TTS model
• espeak-ng: phonemization
• onnxruntime-node: on-device inference

Built by Light Cloud Labs and contributors.