Announcemint

Convert text to speech using AWS Polly, with presets for two-way radio voice prompts and more. Cross-platform desktop app and CLI: one prompt per line, Ogg Vorbis or in-app conversion to WAV.

App name, publisher, docs URL, and GitHub repo are set in brand.json at the project root so you can rebrand in one place (see Rebranding).

Getting the app

Prebuilt installers (Windows MSI and NSIS .exe, macOS .app, Linux AppImage) are published on the Releases page. On Windows, the NSIS installer (.exe) lets you choose per-user or system-wide install; the MSI installs per-machine. You need AWS credentials with Polly permissions before generating speech.

Features

Desktop app

• Main screen: Prompts text area (one per line), output directory, Generate Prompts button. Progress shows the current prompt name, a progress bar, and step-by-step status (Submitted to Polly → Downloading → Converting → Saving). If any destination files already exist, the app asks for overwrite confirmation before generating.
• Status indicator: Green / yellow / red dot next to Settings shows AWS auth and permission status (authenticated + Polly permissions, authenticated only, or not authenticated).
• Help: Opens the docs URL in the system browser (set in brand.json).
• About: App version (including Git SHA for dev builds), license, and project links.

Settings (full-page, expandable drawers)

• Network Proxy: Optional HTTP/HTTPS/SOCKS proxy (protocol, host, port, username, password). Test proxy checks that the app can reach the AWS API through the proxy.
• AWS Account: Credential source (AWS config file or manual keys). Config directory and profile selection; Check credentials and permissions validates auth and reports User ID, Account, ARN, Region, config file path, public IP, and whether polly:DescribeVoices and polly:SynthesizeSpeech are granted.
• Voice Options: Language (including system locale), engine (standard/neural), voice, output preset (e.g. OGG Vorbis, WAV: Two-Way Radio Voice Prompt). Option to remember prompt names after closing.
• Saving Prompts: Output directory and Filename Formatting (e.g. hyphens, underscores, lower/upper case).

Configuration is stored in a platform-specific user config directory (e.g. ~/.config/<publisher>/<appName>/config.json), where publisher and appName come from brand.json.

Output presets

• OGG Vorbis: Polly output as-is.
• WAV: Two-Way Radio Voice Prompt: 8 kHz, 16-bit, mono, little endian. In-app Ogg→WAV conversion (no SoX).

CLI

Same generation logic as the app. Prompts from --file or --text; output directory and optional voice, engine, preset.

# Build first (or use cargo run)
cargo build --release --manifest-path src-tauri/Cargo.toml

# Generate
./src-tauri/target/release/announcemint generate --output-dir ./out --file prompts.txt --preset "WAV: Two-Way Radio Voice Prompt"

Options: --output-dir / -o, --file / -f, --text / -t, --voice-id, --engine, --preset / -p.
Environment: ANNOUNCEMINT_OUTPUT_DIR, ANNOUNCEMINT_VOICE_ID, ANNOUNCEMINT_ENGINE, ANNOUNCEMINT_LANGUAGE_CODE, ANNOUNCEMINT_PRESET, plus standard AWS (AWS_PROFILE, AWS_REGION, etc.).

Requirements

• Node.js and npm (for the frontend and Tauri dev/build)
• Rust (for the backend and CLI)
• AWS credentials with polly:DescribeVoices and polly:SynthesizeSpeech (env vars, ~/.aws config, or manual entry in Settings)

Linux: system libraries for the GUI

The desktop app uses WebKitGTK. Install the development packages so pkg-config can find them (e.g. for just gui or npm run tauri dev).

• Fedora / RHEL:

  sudo dnf install glib2-devel webkit2gtk4.1-devel libsoup3-devel librsvg2-devel
  

  (On some Fedora versions you may need javascriptcoregtk4.1-devel if the above is not enough. If you see gio-2.0 or other GLib .pc not found, install glib2-devel. For AppImage builds, librsvg2-devel is required by the linuxdeploy GTK plugin.)

• Debian / Ubuntu:

  sudo apt-get update && sudo apt-get install -y libwebkit2gtk-4.1-dev libappindicator3-dev librsvg2-dev
  

If the build still reports a missing .pc file:

• Using Linuxbrew or Homebrew on Linux: That pkg-config only searches Homebrew’s prefix, so it never sees system libraries. Prepend the system pkg-config directory so the WebKit/GTK .pc files are found:

  export PKG_CONFIG_PATH="/usr/lib64/pkgconfig:$PKG_CONFIG_PATH"
  

  Then run just gui (or npm run tauri dev) in the same shell. You can add that line to your shell profile if you use the GUI often.

• Otherwise, set PKG_CONFIG_PATH to the directory that contains the .pc file, or install the matching -devel / -dev package for that library.

Linux: dark mode and window theme

On Linux, Tauri does not yet reliably propagate the system theme to the WebView or native window (see tauri#9427). The app works around this by reading the XDG Settings portal (org.freedesktop.appearance.color-scheme) and applying dark/light to the in-app UI. So the content of the window should follow your system dark/light preference. The window border and title bar are drawn by the desktop and may stay light until Tauri/tao adds full Linux theme support; there is no theme option in tauri.conf.json for Linux (it is only implemented on Windows and macOS).

Development

Bootstrap (first-time setup)

Before running the GUI or building the app, set up the repo once:

1. Install dependencies (Node and Tauri CLI):

   npm install
   

   If you see tauri: command not found when running just gui, the Tauri CLI is missing — run npm install from the project root.

2. Create the frontend build output so Tauri can load the UI. The config expects dist/ at the project root. Either build the frontend once:

   npm run build
   

   or create an empty directory (the app will start but the window will have no UI until you run npm run build):

   mkdir -p dist
   

3. On Linux, install system libraries for the GUI and, if you use Linuxbrew, set PKG_CONFIG_PATH as described there.

Then you can run just gui or the commands below.

Daily workflow

• GUI: just gui or npm run tauri dev (Rust is built first so the window opens when the backend is ready).
• CLI: just cli generate --output-dir ./out --text "Hello" or just cli generate --output-dir ./out --file prompts.txt.
• Tests: just test or cargo test --manifest-path src-tauri/Cargo.toml.

Optional: use the Tauri Development launch config in VS Code (Run and Debug) to run the GUI.

Common issues

• just gui fails with "webkit2gtk-4.1", "libsoup-3.0", or "javascriptcoregtk-4.1" not found
  You need the Linux system libraries for the GUI. See Linux: system libraries for the GUI. On Fedora you can run just gui-deps to print the install command.

• frontendDist configuration is set to "../dist" but this path doesn't exist
  Tauri requires the dist/ directory (built frontend). Run npm run build once to create it, or mkdir -p dist for an empty placeholder.

• tauri: command not found
  The Tauri CLI is provided by npm. Run npm install in the project root so node_modules/.bin includes it; then use just gui or npm run tauri dev (do not run tauri directly unless it’s on your PATH).

• Blank window on first open (dev)
  The window can open before the Vite dev server is ready. Reload the window (e.g. Ctrl+R / Cmd+R) or start the dev server first (npm run dev), then run just gui in another terminal.

• Title bar
  The native title bar is drawn by the OS and does not follow the app’s CSS or dark mode. To have the title bar match the app, you’d need to set decorations: false and implement a custom HTML/CSS title bar with a drag region (and optionally window controls via the Tauri window API).

Build

• Windows: On Windows with WiX and NSIS installed, npm run tauri build -- --bundles msi,nsis produces an MSI in src-tauri/target/release/bundle/msi/ and an NSIS .exe in src-tauri/target/release/bundle/nsis/. The NSIS installer prompts for per-user or per-machine install (see bundle.windows.nsis.installMode: "both" in tauri.conf.json).
• macOS (.app): On macOS, npm run tauri build produces the app bundle (e.g. in src-tauri/target/release/bundle/macos/).
• Linux: npm run tauri build produces .deb and AppImage. In environments without FUSE (e.g. Docker, CI), set APPIMAGE_EXTRACT_AND_RUN=1 so the linuxdeploy AppImage can run. If you see “Squashfs image uses (null) compression” when launching via AppImageLauncher, run ./scripts/repack-appimage-gzip.sh path/to/foo.AppImage or run the AppImage directly. Release builds are repacked with gzip so AppImageLauncher can read them. If you get "Could not create default EGL display: EGL_BAD_PARAMETER" and a blank window (e.g. on Fedora with some GPUs): run with WEBKIT_SKIA_ENABLE_CPU_RENDERING=1 ./Announcemint_*.AppImage to force CPU rendering. Release builds use a repack step that injects this into the AppImage’s AppRun (script or binary), so newer builds should avoid the crash; to try GPU again use WEBKIT_SKIA_ENABLE_CPU_RENDERING=0 before launching.

Rebranding

Edit brand.json at the project root:

• publisher – Used in the config directory path (e.g. ~/.config/<publisher>/<appName>/).
• appName – Display name (window title, headers, About). Also used in the config path.
• docsUrl – URL opened by the Help button; also used for the opener permission in capabilities.
• githubRepo – URL for the GitHub link on the About screen.
• shortDescription / longDescription – Bundle metadata (e.g. for installers).

Then run npm run sync-brand to update src-tauri/tauri.conf.json (product name, window title, identifier, CLI description) and src-tauri/capabilities/default.json (docs URL allow list). The build and dev commands run sync-brand automatically. Rebuild the app so the Rust backend picks up the new publisher/app name for the config directory.

Optionally update package.json name and the CLI binary name in src-tauri/src/cli.rs if you want the npm package and CLI command to match your brand.

License

GPL-3.0. See LICENSE.