Let memost write your meeting minutes!

Nov. 1, 2026 · The memost team

Take notes during a meeting and you miss the conversation; record it and you never find time to listen again. AI recording, new in memost 0.41.0, does both at once. Start recording and what people say is written into your note sentence by sentence, with each speaker told apart. When the meeting ends, a draft of the minutes is already in the note.

More than a recording

• Sentence-by-sentence transcription — sentences appear in the note one by one as people talk, each with its start and end time.
• Replay just that sentence — click the time chip at the end of a sentence and only that part of the recording plays, not the whole thing. Checking the transcript takes seconds.
• Speaker labels — when several people talk, sentences are grouped by voice and start with ‘Speaker 1’, ‘Speaker 2’.
• Rename once — click one label and type a name, and that speaker’s labels change everywhere in the note at once.
• The audio stays too — the recording is attached to the note and saved in your space, and ▶ plays it from the start.

Speech recognition and speaker labelling both run on your Mac. Neither the audio nor the text goes out to the internet.

Which models it uses

Speech recognition uses Whisper, the open speech-recognition model released by OpenAI (the recommended size is Large v3 Turbo). Speaker labels use WeSpeaker ResNet34, a model that compares the characteristics of voices. You download each once and they run on your Mac from then on.

One-time setup

Go to Settings ▸ AI connection ▸ AI speech recognition.

1. Pick a Language. If your meetings are all in one language, choose it; if you mix languages, choose Detect automatically.
2. Pick a Whisper model and download it. If you’re not sure, the one marked Recommended is a good start.
3. For meetings with several people, turn on Speaker labels and download the Speaker model.

When you see “Ready — you can use AI recording”, you’re set.

A real example: the weekly marketing meeting

We recorded a one-minute meeting between three people. (For the sample we wrote the conversation, played it through the Mac’s speakers in three synthetic voices, and recorded it with the Mac’s built-in microphone. Every screenshot below shows that result as it came out.)

1. Start with 🎙

Open the note you want the minutes in and click 🎙 (AI recording) in the toolbar above the note. A recording bar appears above the note with the elapsed time, and each finished sentence lands below. Click ⏸ to pause.

2. Stop with ■

When the meeting is over, click ■ in the recording bar. After stopping, memost looks over the whole recording once more to regroup the speakers, then shows “Recognition finished”.

3. Rename the speakers

Click the ‘Speaker 1’ label, type a name and press Enter, and every ‘Speaker 1’ in the note takes that name. Once per person is all it takes.

4. The finished minutes

Now the minutes carry three names. Click a time chip and only that sentence plays; the chip turns orange while it plays. The screenshot shows Daniel’s “The open rate was 42%…” playing.

The spoken “forty-two percent” was written as “42%”. You can also see the one slip in this sample: the one-word “Perfect.” near the end was Emma’s, but it landed on Ryan. More on that below.

Tips

• Stay close to the microphone. Speaker labels work by comparing voices, so voices far from the microphone, people talking over each other, and very short replies like “Yes” or “Perfect.” can end up on the wrong person.
• Recordings don’t get lost. If you lock the space or quit the app while recording, memost stops and saves the recording.
• Let AI tidy the minutes. Hand the transcript to Claude or ChatGPT through the AI connection and get a summary and to-dos by person.

AI recording is available in memost 0.41.0 or later for Mac. Get the latest version from Download.