Transcribe both sides of a meeting on your Mac

An online meeting reaches the Mac through two audio paths. Noats captures the microphone as your channel and system audio as the remote speakers. Whatever your Mac’s microphone hears is attributed to you. The voices arriving from the call are split into per-meeting speaker labels that you can rename.

To transcribe meetings on a Mac without sending audio to the cloud, use Noats on an Apple silicon Mac running macOS 14.2 or later. By default, it captures both paths, transcribes them, separates speakers, writes the note and stores the files locally. Network access covers update checks, the model catalogue and model downloads you approve.

An optional cloud-model request sends only the note and transcript content required for that request directly to your selected provider under its terms. There is no Noats server that receives your meetings. The call receives no added participant from Noats. Zoom, Meet, Teams and the other call services still carry the live conversation under their own terms.

Use Noats on an Apple silicon Mac

Use Noats on a Mac with Apple silicon and macOS 14.2 or later. Give it access to the microphone and system audio when macOS asks. Apple manages microphone access separately from screen and system-audio recording access, including permission to capture audio without capturing the screen.

Local speech recognition alone does not establish where the rest of a meeting goes. Check capture, transcription, speaker separation, note generation and storage separately. Noats performs all five stages on the Mac by default. In Noats, “local” describes the whole meeting path: capture, transcript, speakers, note and files all stay on the Mac by default.

Noats notices when an app takes the microphone and offers to record. Start a note to begin recording. Your microphone channel is assigned to you while the system-audio channel supplies the other voices.

You can name each separated speaker during the conversation or afterward. We keep those names scoped to that meeting. Our explanation of how Noats separates and names speakers covers that review process in more detail.

Prove the workflow with the network disconnected

Download the models you approve before running the test. Transcription and speaker separation run on Apple silicon, so a meeting with the network disconnected is processed the same way as one at your desk once those models are present. The following test exercises the microphone channel, the system-audio channel, speaker separation and note generation together.

1. Disconnect the Mac from the network after confirming that the approved models are downloaded.
2. Create a note in Noats to begin recording.
3. Play a saved speech recording through system audio rather than using a streaming source.
4. Speak into the microphone while recording continues so both audio paths contain speech.
5. Allow the local models to finish processing, then confirm that both sides, separate speaker labels and the written note appear in Noats.

A successful run shows that capture, transcription, speaker separation and note writing can complete without a network connection. It does not audit behavior while the Mac is connected. For that check, inspect network activity during an ordinary recording and compare it with the paths the app describes.

By default, network activity is limited to update checks and downloads, loading the model catalogue when you open Models and model downloads you approve. Automatic updates can be switched off in Settings under About. The app contains no analytics and no telemetry.

If you select a cloud model, the required note and transcript content goes directly to that provider. Our guide to local notes and optional provider requests explains each connection.

Keep the files and review the names

Speaker labels remain editable during and after the meeting. Review each label against the transcript, assign the names you want and correct the note before exporting it. Separation is scoped to the current meeting, so the labels represent the voices in that recording and the names you supplied for it.

Your audio, transcripts, speaker labels, notes, folders and chat history are written to a database and plain files in your own Application Support folder. Chat history is scoped to one meeting. Any note can be exported as Markdown. These are user-owned files in a standard per-user location rather than records held by a Noats meeting service.

Apple’s Voice Memos transcription covers live and completed recordings. Its recording instructions list the built-in microphone, a supported headset and an external microphone as inputs. For a spoken memo from one listed input, Voice Memos is the more direct choice. A two-sided online meeting needs the separate microphone and system-audio capture described above.

Apply the same inspection to any local transcription app. Identify where each audio path is captured, where speech becomes text, where speakers are separated, where the note is generated and where the resulting files are retained. Then inspect update traffic, model downloads and optional provider requests as separate network paths.

Zoom, Google Meet, Microsoft Teams and other call services still carry the call under their own terms.