RegardingWork Voice

Your meetings. Recorded and transcribed on your Mac.

Capture your microphone and Mac system audio as separate tracks, then create a private, on-device transcript—all from a visible menu-bar recorder.

Free for Macs running macOS 15 or later. Apple silicon recommended. Get the latest build on GitHub Releases.

Local-first by design

RegardingWork Meetings records both sides of a Mac meeting and creates a local transcript without sending the conversation to a transcription cloud.

On-device transcription

Parakeet transcription runs locally after the speech model is prepared.

No cloud transcription

Meeting audio and transcripts are not uploaded to RegardingWork or a third-party service.

Original audio preserved

Separate microphone and system-audio tracks remain available for review or retranscription.

Visible recording health

The menu shows whether each track is active, silent, stalled, or failed.

No telemetry

The app does not transmit analytics, usage events, meeting metadata, or transcript text.

How it works

Start, confirm both tracks, stop—and keep original audio plus a local transcript.

  1. Start from the menu bar

    Choose Start recording from the Meetings waveform icon. Recording is never hidden: the icon changes, macOS shows its recording indicator, and the menu displays elapsed time.

  2. Confirm both tracks

    Meetings records the default microphone and all audio played by the Mac as separate local tracks. The menu shows the health of each track so a silent or failed source is visible during the meeting.

  3. Stop and transcribe locally

    Choose Stop recording. The app keeps the original audio and runs Parakeet on the Mac, producing a readable Markdown transcript and structured JSON.

Why teams choose Meetings

Keep meeting data on your Mac

Audio, transcript text, timing metadata, and recovery files remain in a local recordings folder. There is no RegardingWork meeting cloud, account, or transcript dashboard.

Capture both sides cleanly

The microphone track is the local side of the conversation; the system track is audio played by the Mac. Separate tracks improve reviewability and give the transcript simple source labels.

See recording problems while they matter

The menu reports recent signal and buffer health for both sources. A track that is silent, stalled, or failed is not presented as healthy.

Recover more after an interruption

An in-progress manifest is written while recording. After an unexpected exit, the app can discover sessions with readable audio and mark them as recovered rather than silently discarding them.

Keep an auditable source

Each session retains original CAF audio plus structured metadata. Suspected speaker-playback echo is handled conservatively while preserving original audio and canonical JSON.

Work without a meeting bot

Meetings captures audio locally from the Mac. It does not join the call as a participant, invite a bot, or send the meeting to an external AI service.

Suggested use cases

  • Keep a local record of customer or partner calls
  • Review technical discussions and project decisions
  • Create searchable source material for meeting notes
  • Revisit exact wording and timestamps after a long conversation
  • Preserve research interviews with appropriate consent
  • Private records when cloud transcription is not allowed

Two source tracks—not every speaker

RegardingWork Meetings labels microphone speech as me and Mac playback as them. These labels describe where the audio came from. They are not voice identification and not true multi-speaker diarization.

If several remote participants speak through the same meeting app, their speech shares the system track. The app does not claim to identify which remote person spoke.

Headphones are preferred because they reduce system playback bleeding back into the microphone.

The meeting stays local.

RegardingWork Meetings stores audio and transcripts on the Mac and transcribes them on-device. It does not upload recordings, transcript text, metadata, or logs. It does not include analytics, telemetry, cloud summarization, external AI processing, SSO, or synchronization.

The first transcription-model preparation requires an internet connection to download the local Parakeet model. After preparation, transcription runs on the Mac.

Important: the system-audio track is deliberately broad—it records everything the Mac plays during the session, including notifications, music, browser tabs, and unrelated applications. Enable Focus mode, close unrelated media, and follow applicable recording consent requirements.

What each session contains

By default, sessions are stored in ~/RegardingWork/Meetings. The app does not automatically upload, expire, or delete them.

mic.caf

Local microphone audio

system.caf

Everything played by the Mac

meta.json

Timing, track health, attribution, recovery status

transcript.md

Readable transcript with visible warnings

transcript.json

Canonical timed transcript

transcribe.log

Progress and errors (without transcript text)

Product requirements

  • Mac running macOS 15 or later
  • Apple silicon recommended for transcription performance
  • Microphone permission
  • Screen & System Audio Recording permission
  • Internet connection during initial speech-model preparation
  • Headphones recommended for the cleanest two-track result

FAQ

Does my meeting audio leave my Mac?

No. RegardingWork Meetings records and transcribes locally. It does not upload audio or transcripts to RegardingWork or a cloud transcription provider.

Does it record the other people in the meeting?

It records audio played by the Mac, which normally includes remote meeting participants. The system track also captures unrelated Mac audio such as notifications, music, and other applications.

Does it identify individual speakers?

No. “me” means the microphone track and “them” means the Mac system-audio track. This is source attribution, not individual voice identification or multi-speaker diarization.

Why are headphones recommended?

When meeting audio plays through speakers, the microphone can hear it again and create duplicate-looking speech across both tracks. Headphones keep the tracks cleaner.

What happens if a recording is interrupted?

The app maintains an in-progress session manifest and is designed to discover readable audio after relaunch. Recovered sessions are marked clearly, and missing or unreadable tracks are reported rather than fabricated or discarded.

Can I tell if a track is silent?

The menu reports microphone and system-track health during recording. It distinguishes active signal from starting, silent, stalled, and failed states.

Does it summarize meetings or create action items?

No. The current product records and transcribes locally. It does not send the transcript to an external AI system or automatically generate summaries, decisions, or action items.

Which languages are supported?

The current shipped Parakeet v2 transcription engine is English-only. Multilingual options are future possibilities, not current features.

Is an internet connection required?

An internet connection is needed for the initial on-device speech-model download. Once the model is prepared, recording and transcription run locally.

How long are meetings retained?

There is no automatic retention or cleanup policy. Local meeting files remain until you delete them. Finder Trash, backups, snapshots, exports, and synced copies must be managed separately.

Do I need permission to record other people?

Recording-consent requirements vary by location, organization, contract, and meeting platform. Notify participants and obtain consent where required before recording.

Keep the conversation. Keep it local.

Record both sides of your Mac meetings and create a private, on-device transcript without adding a meeting bot or sending the conversation to a transcription cloud.

Download for Mac

Open-source and free to download. Grab the latest release from GitHub Releases, then follow the app’s permission and consent guidance on first run.

Part of RegardingWork Voice

Meetings preserves local recordings and transcripts. Need push-to-talk text insertion instead?