Local-first voice notes for Windows. Press a hotkey, speak, release. Get a transcript — your way.

Phoneme bridges the gap between quick voice dictation and your personal knowledge management systems. It is designed for power users who want the friction-free experience of hitting a hotkey to capture a thought, but without the privacy concerns, subscription fees, or cloud lock-in of modern AI tools.

Everything runs 100% locally on your machine.

When you press your global hotkey (e.g., Ctrl+Alt+Space), Phoneme records your voice. When you stop, it leverages a local Whisper instance to transcribe your speech into text. Finally, it pipes that text through your own scripts (hooks) or into an LLM (like Ollama) for cleanup, formatting, or translation.

The app does not force you into a specific ecosystem. It transcribes. You decide where it goes.

This is the stable public release. Everything in this list is available today:

• Press-to-talk & toggle modes — bind any global hotkey, choose Hold or Toggle behaviour.
• Local Whisper transcription — uses your own whisper-server or the bundled one; your audio never leaves your machine.
• AI post-processing — optionally clean up, format, or translate transcripts through a local Ollama model or OpenAI. Includes 9 preset prompts.
• Hook pipeline — every transcript is delivered to your script as JSON on stdin. Chain scripts, POST to webhooks, send to Obsidian, Org-mode, Notion, or anywhere. Five reference hooks included.
• Full CLI — every GUI action is also available as phoneme commands. Works with AutoHotkey, Kanata, Stream Deck, or any hotkey daemon.
• Tags — attach colour-coded tags to recordings; filter and search the catalog.
• Transcript editor — edit transcripts in-app with optional full Vim mode (visual, linewise, mouse selection all work).
• Doctor — built-in health checker that tests the daemon, audio dir, hooks, Whisper server, and Ollama, with one-click fixes.
• 11 themes — Catppuccin Mocha/Macchiato/Latte, Dracula, Everforest, Gruvbox, Nord, One Dark, Rosé Pine, Solarized Light, Tokyo Night.
• Auto-updater — downloads and installs new releases directly from GitHub.
• Export — bulk export all recordings and metadata as a zip archive.

Download the latest .msi from the releases page and run it.

On first launch, the wizard walks you through:

• Pointing at your whisper-server (or using the bundled one with your GGUF model)
• Picking your microphone
• Picking your hook script (default writes to stdout)
• Setting your global hotkey
• Choosing your aesthetic theme

Requirements: Windows 10/11. A locally running whisper-server (installed alongside Phoneme in bundled mode, or run separately in external mode). For bundled mode, you also bring your own GGUF model file (e.g., ggml-base.en.bin).

No cloud. No telemetry. No update pings. The only network calls Phoneme makes are to your configured whisper-server endpoint, your chosen local LLM, and (optionally) Hugging Face when you explicitly click to download a model during setup. Your voice and your thoughts stay on your hard drive.

Because Phoneme pipes JSON directly into your own scripts (hooks), it pairs perfectly with local-first, text-based productivity apps:

• Obsidian: Write a hook that automatically appends your transcript to your daily note.
• Logseq / Roam Research: Format your transcript as a bullet point and append it to your journal file.
• Emacs (Org-Mode): Pipe the output directly into org-capture.
• Notion: Use a Python or PowerShell script to POST the JSON payload to the Notion API.

If Phoneme doesn't quite fit your workflow, or if you're on a different operating system, check out these excellent alternatives:

• MacWhisper & Superwhisper: Fantastic, highly polished local dictation apps built exclusively for macOS.
• AudioPen: A popular cloud-based web app that records and beautifully summarizes your thoughts.
• AquaVoice: A voice-native text editor.

Every action available in the GUI is available from the command line:

phoneme record --oneshot                        # record + transcribe + print
phoneme record --start                          # non-blocking start
phoneme record --stop                           # non-blocking stop
phoneme list --since 2026-05-19                 # query the catalog
phoneme show 20260519T143500823                 # one recording's details
phoneme export backup.zip                       # bulk export audio and metadata
phoneme doctor                                  # health check
phoneme config reload                           # hot reload config from disk
phoneme watch                                   # subscribe to events as JSON

We deliberately built Phoneme with a CLI-first architecture to provide you with the exact flexibility that big tech products won't. You aren't locked into our built-in global hotkeys. Advanced users can bind phoneme CLI commands to any hotkey daemon, window manager, or macro pad they prefer—whether that's AutoHotkey, Kanata, WHKD, or a Stream Deck.

Because the CLI seamlessly controls the daemon, setting up a custom workflow is as simple as making your tool shell out to phoneme record --start and phoneme record --stop!

A hook is your script. Phoneme invokes it with the transcript as JSON on stdin. Ship your own or use one of the four reference hooks:

Chain multiple hooks under [hook] commands = ["script1.ps1", "script2.bat"], and optionally POST the JSON payload to a webhook_url simultaneously.

See docs/hooks.md for the full contract.

Three binaries, three libraries, one workspace:

                            ┌──────────────────────────────────┐
                            │          phoneme-daemon          │
                            │ (Headless: audio, queue, catalog)│
                            └───────────────▲──────────────────┘
                                            │
                      named pipe (\\.\pipe\phoneme-daemon)
                                            │
             ┌──────────────────────────────┴──────────────────────────────┐
             │                                                             │
             ▼                                                             ▼
    ┌─────────────────┐                                           ┌─────────────────┐
    │     phoneme     │                                           │  phoneme-tray   │
    │      (CLI)      │                                           │   (Tauri GUI)   │
    └─────────────────┘                                           └─────────────────┘

# Requirements: Rust 1.75+, Node 20+, pnpm 9+, tauri-cli 2

cd frontend && pnpm install && cd ..
cargo install tauri-cli --version "^2.0" --locked
cargo tauri build

The MSI lands at target/release/bundle/msi/.

For development (with hot reload):

# Terminal A
cargo run -p phoneme-daemon -- --foreground

# Terminal B
cargo tauri dev

For a detailed look at our upcoming features—including macOS/Linux ports, Streaming Transcription, and Bundled Ollama—please see our Roadmap.

We welcome contributions! If you're interested in helping improve Phoneme, please check out our Contributing Guide to learn how to set up the development environment, build the app, and submit pull requests.

Have a question or idea? Start a discussion.

MIT OR Apache-2.0.

Phoneme is built by @namefailed. It is not a commercial product, has no telemetry, and never will.