TTP — Talk To Paste

Press a shortcut. Speak. Your words appear wherever you're typing.

Website · Download · Changelog

---

TTP is a lightweight desktop app that turns speech into text in any app. Hold a shortcut, speak, and your words are transcribed and pasted wherever your cursor is. No switching apps, no copy-pasting, just talk.

Built for long dictations: minutes of speech, with pauses to think, come back whole.

Free. No account required. Bring your own Groq API key.

Features

• Accurate on long dictations — Whisper large-v3 on Groq. Pauses and quiet speech are kept, not dropped as silence.
• Works everywhere — Paste into Slack, VS Code, Gmail, Notion, any app where you type.
• AI polish — Removes filler words and fixes grammar with a Groq LLM pass.
• Smart dictionary — Learns your names, jargon, and technical terms post-paste.
• Any shortcut — Fn, a key, a modifier or a mouse button, plus an optional second one. Hold to talk, or double-tap for hands-free.
• Auto-stop after silence (opt-in) — End the recording automatically when you've stopped speaking.
• Mac + Windows — Native app, lives in your menu bar / system tray. Speaks English and French.
• Clear about your data — Your voice goes to Groq for transcription, with your own key. There is no TTP server. Keys live in your OS keychain, history stays local, crash reports are opt-in.
• Updates without interruption — Minisign-verified updates download in the background and install during a two-minute pause.

Trust & Security

TTP runs on your machine and handles your voice and your API keys, so we treat the security pipeline as a feature.

• Signed + notarized macOS builds — every release is code-signed with an Apple Developer ID certificate and notarized through Apple's notary service before shipping.
• Code-signed Windows installers — .msi and NSIS -setup.exe artifacts are signed in CI when the Windows certificate is configured.
• Minisign-verified auto-updates — the Tauri updater verifies every update payload against a minisign public key embedded in the app at install time; tampered updates are rejected.
• OS keychain for API keys — your Groq API key lives in the macOS Keychain (and Windows Credential Manager on Windows), not in plaintext config files.
• HMAC-signed local caches — the offline licence cache (for the optional supporter licence) is HMAC-signed with a per-machine secret stored in your OS keychain, so a licence file forged on one machine won't be accepted on another.
• Telemetry off by default — Sentry crash reporting is opt-in and disabled until you explicitly enable it in Settings. No usage analytics are collected.
• Strict CSP, no remote webview content — the embedded webview only loads bundled assets; the only outbound endpoints reachable from the renderer are Groq and Sentry.
• Third-party calls are listed in the privacy policy — Groq, Lemon Squeezy, GitHub (updater feed), Sentry. No other network calls are made.

Found a security issue? Please report it privately — see SECURITY.md. The threat model that scopes "in scope" vs "out of scope" reports lives in THREAT_MODEL.md.

How It Works

1. Hold your shortcut — Fn by default on Mac; any key, modifier or mouse button you choose.
2. Speak — Talk naturally. TTP records your voice in the background.
3. Text appears — Your words are transcribed and pasted into the active app, typically about a second and a half after you let go.

Installation

Download

Grab the latest release for your platform:

• macOS — .dmg from Releases
• Windows — -setup.exe or .msi from Releases

Setup

1. Install the app
2. Get a free API key from Groq Console
3. Paste the key in TTP's setup screen
4. Start talking

Build from Source

Prerequisites

• Node.js 18+
• Rust (latest stable)
• Tauri CLI

Steps

# Clone the repo
git clone https://github.com/AmirK-S/TTP.git
cd TTP

# Install dependencies
npm install

# Run in development
npm run tauri dev

# Build for production
npm run tauri build

Tech Stack

Layer	Technology
Framework	Tauri 2
Frontend	React + TypeScript
Backend	Rust
Styling	Tailwind CSS
Transcription	Groq Whisper (whisper-large-v3)
AI Polish	Groq LLM (openai/gpt-oss-120b)
Landing Page	Astro

Project Structure

TTP/
├── src/                  # React frontend
│   ├── components/       # UI components + design primitives (Button, Modal, ...)
│   ├── hooks/            # Recording / transcription / updater / Tauri events
│   ├── lib/              # Pure helpers (theme, i18n, PII scrub, ...)
│   ├── stores/           # Zustand settings store
│   ├── windows/          # Settings · Onboarding · FloatingBar · PermissionHelper
│   └── i18n/             # EN + FR translation tables
├── src-tauri/            # Rust backend
│   └── src/
│       ├── transcription/  # Whisper + LLM polish + cleanup + dictionary apply
│       ├── dictionary/     # Smart dictionary with auto-detection
│       ├── audio_capture.rs / audio_monitor.rs / vad.rs   # cpal + RMS + VAD
│       ├── fnkey.rs / fnkey_fsm.rs  # macOS Fn key (FSM is pure + unit-tested)
│       ├── state.rs / shortcuts.rs / tray.rs              # Recording lifecycle
│       ├── settings/ + usage/ + history/ + licensing/     # Persistence layers
│       └── telemetry/ + logging.rs                        # Observability
├── docs/architecture.md  # Runtime architecture + diagrams
├── THREAT_MODEL.md       # In-scope vs out-of-scope security boundaries
└── landing/              # Astro landing page (ttp.amirks.eu)

Running tests

• Frontend — npm run test:run (vitest + happy-dom, ~150 tests).
• Rust — cd src-tauri && cargo test --lib (~470 tests across the Fn FSM, HMAC, settings, pipeline classifiers, silence gate, dictionary, VAD).
• Build-time gates — npm run i18n:check validates locale parity AND that every t('xxx') callsite resolves against a real key. Wired into npm run build.

Author

Amir Kellou-Sidhoum — AI Engineer / Builder / Consultant

• LinkedIn
• Website

---

Made with care.