Every Plaud recording,on your Mac,as text.

My Plaud is a menu bar app. It keeps the audio of every recording in your Plaud account in a folder you own, transcribes it and tells the speakers apart. A language model can clean the transcript, name the recording and write a summary.

Download for macOSVersion 0.1.0, for macOS 27 or later. Free.
A recording in My Plaud: its name, three speakers, a player bar and the transcript with a speaker and a time on every turn

It keeps a copy of everything

Every 15 minutes, at launch and when the Mac wakes, the app downloads what is new. Each recording gets its own folder, named by date and time, with the original audio in it.

Nothing is ever changed in the Plaud cloud, and nothing in the archive is deleted for good. A recording removed in Plaud stays, and Delete moves a folder to the macOS Trash.

~/Plaud/2026/11/2026-11-19_1455 Website relaunch kickoff/

audio.oggThe original audio
transcript.mdWho said what, and when
summary.mdThe summary
speaker-1.oggA voice sample of a speaker without a name
video.movThe screen, when you recorded it on the Mac

You choose what transcribes it

Use a hosted engine with your own key, or Whisper on the Mac itself. The transcript is a plain Markdown file next to the audio.

EngineYou needTells speakers apartWhere the audio goes
ElevenLabs ScribeAn API keyYes, and it recognises the voices in your speaker libraryTo ElevenLabs
OpenAIAn API keyWith its diarising modelTo OpenAI
OpenRouterAn API keyDepends on the modelTo OpenRouter and the provider that serves the model
Whisper on this MacA 1.7 GB downloadYes, in each recordingNowhere. It stays on the Mac

It knows who is speaking

Name a speaker once and the name appears in the transcript, in the list and in the summary. With ElevenLabs, the app cuts a voice sample of every unknown speaker, so later recordings recognise them by themselves.

The Recordings pane: a table of eleven recordings with date, name, duration, speakers and state

A model cleans it up, names it and sums it up

This step is optional. A language model fixes punctuation and mis-heard words, removes filler, gives the recording a short name and writes a summary. Speaker names and times never pass through the model, and the original transcript is kept.

Use a model from Anthropic, OpenAI or OpenRouter, or a model server on your Mac such as LM Studio, where the text never leaves the machine.

The Summary tab of a recording, with the headings Overview, Decisions, Action items and Open questions

It records on the Mac too

Start recording in the menu records the microphone and everything the Mac plays, so both sides of a call in any app end up in one file, with or without headphones.

Add a video of the screen or of one window. The recording lands in the same archive and is transcribed, named and summarised like one from Plaud.

A recording made on the Mac: the video of the shared window above the player bar and the transcript

Your recordings stay yours

Plaud receivesYour login, and requests for your recordings. The app never changes anything in the Plaud cloud.
The transcription service receivesThe audio file, the language and your key terms. No title, no folder name, no recording ID.
The model provider receivesThe transcript text, for the cleanup and the summary. Never the audio.
Everyone else receivesNothing. There is no My Plaud server, no account and no telemetry.

With Whisper on this Mac and a local model, the only network contacts are the downloads from Plaud and the one-time download of the speech model. The archive is ordinary files that any other program can read.

Read exactly what is sent

Get My Plaud

Open the disk image, drag the app to Applications and log in to Plaud. The first sync downloads every recording.

Download for macOSVersion 0.1.0, for macOS 27 or later. Free.