VoicePen

VoicePen is a small offline macOS app for dictation, voice sessions, and meeting recording.

I built it for myself because I wanted something simple, comfortable, and free: quick push-to-talk dictation, a normal sessions/history flow, and a meeting recording mode with local transcription and experimental speaker diarization. I could not find that exact mix without subscriptions, cloud transcription, or too much product ceremony, so I decided to make it and keep the project in an open-source spirit.

It is still early and personal, but useful enough to share.

Features

• Push-to-talk dictation: hold the hotkey, speak, and VoicePen inserts the final text into the active app.
• Sessions: keep local history for ordinary dictation sessions.
• Meetings: record microphone plus system audio, transcribe locally, and optionally add timecodes and speaker labels.
• Custom dictionary: normalize names, product terms, and recurring recognition mistakes.
• Local-first privacy: no cloud transcription, no analytics, and no runtime data collection.

Status

VoicePen is an early macOS build for my own workflow and a small Friends & Family circle. Expect rough edges, but also a bias toward practical fixes over big platform ideas.

Requirements

• macOS 15 or newer on Apple Silicon.
• Microphone permission.
• Accessibility permission for text insertion and global hotkeys.
• System Audio Recording permission for Meeting Mode.
• Xcode installed for local development.

Install

If you received VoicePen.app or VoicePen.zip:

1. Unzip the file.
2. Move VoicePen.app to /Applications.
3. Right-click VoicePen.app.
4. Click Open.
5. Confirm opening the app if macOS warns that it is from an unidentified developer.

Use right-click -> Open on the first launch. Unsigned development builds can be blocked by Gatekeeper; if that happens, run:

xattr -dr com.apple.quarantine /Applications/VoicePen.app
open /Applications/VoicePen.app

Build Locally

make resolve-packages
make build
make run

Run tests:

make test

Development tooling lives in Docs/development.md, and test guidance lives in Docs/testing.md.

Development Workflow

VoicePen uses a spec-driven workflow for product behavior. Before changing how the app behaves, update or create the relevant spec in Specs/.

The short version:

1. Find the relevant spec in Specs/index.md.
2. Update acceptance criteria before implementation.
3. Add or update focused tests for the changed behavior.
4. Implement the smallest useful change.
5. Run make test before handoff for production behavior changes.

Presentation-only and README changes do not need a spec.

Local Data

VoicePen stores downloaded models here:

~/Library/Application Support/VoicePen/Models/

VoicePen stores settings, history, dictionary entries, and meeting transcripts in a local SQLite database:

~/Library/Application Support/VoicePen/VoicePen.sqlite

Optional proxy settings for model downloads can be configured here:

~/.voicepen/config.toml

Support

VoicePen is free because I wanted this workflow to exist without another subscription. Feedback, issues, and pull requests are welcome. I may add a "buy me a beer" link later for people who want to support the work, but the app should stay useful without paying for it.

Maintainers

• Release and signing notes: Docs/releasing.md
• Architecture decisions: Docs/adr/
• Product specs: Specs/