Analysis updated 2026-05-18
Dictate text directly into a terminal, code editor, or browser without typing
Get fast, fully offline speech-to-text with no data sent to a remote service
Use custom vocabulary so technical terms like tool names transcribe correctly
| jaswanth1524/vox-terminal | 0xallam/my-recipe | 0xhassaan/nn-from-scratch | |
|---|---|---|---|
| Stars | 0 | — | 0 |
| Language | Python | Python | Python |
| Last pushed | — | 2022-11-22 | — |
| Maintenance | — | Dormant | — |
| Setup difficulty | moderate | moderate | moderate |
| Complexity | 2/5 | 2/5 | 4/5 |
| Audience | developer | general | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires Apple Silicon and macOS 13+, about 3GB free disk space, and granting microphone, accessibility, and input monitoring permissions.
Vox Terminal is a dictation app for Apple Silicon Macs that lets you speak instead of type into whatever text field you are currently focused on, whether that is a terminal, a code editor, a browser, or any other app. You hold the Right Option key, speak, release the key, and the words you said are pasted in as text within about four tenths of a second. All the speech recognition happens locally on your own machine using the MLX framework, with no audio ever sent to a remote transcription service. The app lives in your menu bar rather than as a normal window, and its icon changes to show whether it is ready, recording, or transcribing. From the menu you can also open settings, run diagnostics, view recent transcript history, and see performance reports, though history and audio are kept only in memory for the current session and cleared when the app quits. Getting it running requires macOS 13 or newer on an Apple Silicon Mac, with about 3 gigabytes of free space for the app and its chosen speech model. You can either download a signed release archive and verify its checksum before installing, or build it from source using Python, a tool called uv, and a small set of Homebrew packages. After installing, you approve a one time model download and grant the app three specific macOS permissions: microphone access to record audio, accessibility access to paste the finished text, and input monitoring so it can detect the global hotkey. By default it uses a fast English speech model called Parakeet, with an optional multilingual Whisper model for other languages or higher accuracy, and settings such as custom vocabulary words or whether to hold or toggle the recording key can be adjusted in a configuration file. Aside from the initial model download you approve, the app has no telemetry, analytics, or ongoing internet connection, and blocks outbound network activity by default during normal use.
A local, offline dictation app for Apple Silicon Macs that pastes your speech as text into any focused app when you hold a hotkey.
Mainly Python. The stack also includes Python, MLX, Parakeet.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.