wordbird 0.3.3

Voice dictation daemon using NVIDIA Parakeet on Apple Silicon

🦜 Wordbird

Contextual voice dictation for macOS. Powered by NVIDIA Parakeet running locally on Apple Silicon via MLX.

Press a hotkey, speak, and your words are transcribed and pasted into whatever app is focused. A small LLM post-processes the transcription to fix errors, using project-specific context from a WORDBIRD.md file.

Getting started

Requires macOS on Apple Silicon (M1+) and Python 3.10+.

# Run with uvx (no install needed)
uvx wordbird

# Or run in the background
uvx wordbird start
uvx wordbird stop
uvx wordbird status

Context-aware correction

The key idea behind wordbird is contextual transcription correction. When dictating into Terminal.app, wordbird detects the focused tab's working directory and looks for a WORDBIRD.md file up the directory tree. This lets you teach wordbird your project's domain:

Context detection works with:

• Terminal.app — detects the focused tab's shell working directory
• VS Code / VS Code Insiders — via the Wordbird extension, which works with local and remote (SSH) workspaces

Transcription and pasting work in any app.

uvx wordbird init

This creates a WORDBIRD.md with the default prompt template. Edit it to add your project's terms, names, and jargon:

---
transcription_model: mlx-community/parakeet-tdt-0.6b-v2
fix_model: mlx-community/Qwen2.5-1.5B-Instruct-4bit
---

Fix transcription errors. Output only the corrected text.

Example 1:
Input: "the java script function isnt working"
Output: "The JavaScript function isn't working."

Example 2:
Input: "check the get ignore file for the repo"
Output: "Check the .gitignore file for the repo."

Example 3:
Input: "we need to refactor the a p i endpoint"
Output: "We need to refactor the API endpoint."

Key terms: MyClass, some_function, PostgreSQL
Names: Alice, Bob

Input: "{{ transcript }}"
Output:

The file is a Jinja template. {{ transcript }} is replaced with the raw transcription. If omitted, the transcript is appended automatically.

The YAML front matter lets you override models per-project. When you dictate into a Terminal tab whose shell is in that directory (or a child), wordbird picks up the nearest WORDBIRD.md and uses it.

Hotkeys

Action	Default
Toggle recording	Right ⌘ + Space
Hold to record	Hold Right ⌘ for >1s, release to transcribe

Hotkeys are configurable:

--hold-key KEY       Hold key (default: rcmd). Options: rcmd, lcmd, ralt, lalt, rshift, lshift, rctrl, lctrl
--toggle-key KEY     Toggle key (default: space). Options: space, return, tab, escape

Options

--model MODEL        Transcription model (default: mlx-community/parakeet-tdt-0.6b-v2)
--fix-model MODEL    Post-processor model (default: mlx-community/Qwen2.5-1.5B-Instruct-4bit)
--no-fix             Disable LLM post-processing

Dashboard

Wordbird runs a local web dashboard at localhost:7870. Click the bird in the menu bar → Dashboard… to open it.

• 📝 History — browse all your transcriptions with timestamps, app name, working directory, and duration. See both the original and corrected text.
• ⚙️ Settings — configure hotkeys, models, and post-processing. Changes take effect immediately without restarting.

You can also view history from the command line:

uvx wordbird history

Menu bar

Wordbird shows a bird icon in the menu bar:

• ⚪ White — idle
• 🟡 Yellow — connecting mic
• 🔴 Red — listening
• ✨ Sparkles — transcribing

Permissions

Wordbird needs three macOS permissions, granted to your terminal app:

• 🎤 Microphone — to record your voice
• 🔐 Accessibility — to paste text and intercept the hotkey
• ⌨️ Input Monitoring — to detect the global hotkey

Wordbird checks these on startup and tells you what's missing.

License

MIT