Skip to content
Book a demo Try the demo MacWindows

Voice note widget

Record a voice or video note on a desk. PlexiDesk transcribes it, offers a summary, a cleaned-up or speaker-labelled version, and suggests next steps.

All plansUpdated

The Voice note widget lets you talk instead of type. Press Voice or Video, say what is on your mind, and press Stop. PlexiDesk turns the recording into text and shows it as a short summary, a cleaned-up transcript, a transcript split by speaker, or the full transcript, whichever you choose. It also suggests things to add to your workspace, such as a to-do list or a page, and you decide which of them to keep.

Use it to capture an idea while it is fresh, to debrief straight after a call, or to talk a problem through when writing feels like too much. The recording stays with the note, so you can play it back at any time.

More about this widget Coming soon

Before you start

  • The Voice note widget is on every plan.

  • You need a desk open to add one.

  • PlexiDesk needs your microphone, and your camera too for a video note. A Mac asks for permission the first time you record. See Allow microphone, camera and screen recording on a Mac.

  • Turning speech into text is called transcription, and you choose how it happens in Settings > AI, under Voice · transcription provider:

    • Cloud sends the recording to OpenAI with your own OpenAI key, which you add in the same tab under OpenAI API key. New installs start on Cloud.
    • Local transcribes on your computer and needs no key. The first time you choose it, PlexiDesk downloads a small speech model, about 80 MB.

    See Choose how speech is transcribed: Cloud or Local Coming soon.

  • The Summary, Cleaned and Speakers views and the suggested actions use your own Anthropic key, which you add in Settings > AI under Anthropic API key. This applies even if you use PlexiDesk AI credits for everything else. The Full view needs no Anthropic key. See Add your own API keys.

To open Settings, click the gear at the right end of the header, then the AI tab.

Add a Voice note

  1. Open the desk you want the note on.
  2. Click Add widget in the centre of the header.
  3. Click Voice note.

The Add widget menu groups widgets into columns by what they are for. Voice note is in the Write & capture column, with Sticky, Note, Markdown, Page, Card and Living Doc. You can also type voice in the search box at the top of the menu and press Enter to add it.

A new Voice note is 320 × 240. It shows a microphone, the line Record → AI transcribes, summarises, suggests tasks., and two buttons: Voice and Video.

Other ways to add one

  • Command palette. Press ⌘ K on a Mac or Ctrl K on Windows, type voice and choose Add Voice note.
  • Tool rail. Rest the pointer on the tool rail at the right of the desk so it opens, then click its Add widget button. Voice note is in the Tools group. See The tool rail on the right of a desk.
  • Right-click the desk. Right-click an empty part of the desk and choose Add object > Tools > Voice note. The note appears where you clicked.
  • From Home. The Transcribe tile on Home records straight away and then asks where to keep the result. Choose On a desk to save it as a Voice note on the desk you pick. See Transcribe widget.

Voice note has no single-key shortcut.

Record a note

  1. Click Voice to record sound only, or Video to record from your camera as well.
  2. If your computer asks whether PlexiDesk may use the microphone (and the camera, for video), allow it.
  3. Speak. A timer counts up beside a pulsing red microphone (a camera for video). For a video note, your camera picture shows above it.
  4. As you talk, live captions appear underneath. They are a preview to show the note is hearing you. They are not saved.
  5. Click Stop when you have finished.

PlexiDesk saves the recording first, then shows Transcribing audio… while it turns the recording into text. Next it shows Cleaning up the text… (or Identifying speakers… on the Speakers view) while it prepares the view you are on and looks for suggested actions. The note then shows the text, with a player underneath to replay the recording.

A video note's title bar reads Video note instead of Voice note.

Choose how the text is shown

The four buttons along the top of a finished note change how the text is shown. The one you are on is highlighted.

  • Summary (the starting view) gives a short paragraph about what you said, followed by a Key points: list.
  • Cleaned is your transcript with filler words such as "um" and "you know", false starts and repeated phrases taken out. Your meaning and the order you said things in stay the same, and punctuation and paragraph breaks are added.
  • Speakers splits the transcript into turns labelled Speaker 1, Speaker 2 and so on, up to four. The turns are worked out from the conversation itself, such as a question followed by its answer, rather than from people's voices. If one person talks throughout, the whole note is labelled Speaker 1.
  • Full is the transcript exactly as it came back, with no AI changes.

Full appears instantly. Summary, Cleaned and Speakers are written by AI each time you switch to them, so allow a moment. The note remembers the view you last chose, and the next recording you make in it opens on that view.

Beside the buttons, the note shows how long the recording is. With Cloud transcription it also shows the language it detected. At the bottom right, it counts the characters in the text on screen.

Act on what you said

If you mentioned things you want done, a section called AI proposed actions appears under the text, with the number of suggestions beside it. There can be up to eight. Each card is one suggestion, with the reason for it, for example:

  • a new desk for a piece of work you described,
  • a to-do list of the steps you mentioned,
  • a Sticky, Note, Markdown or Page widget holding some of what you said.

Nothing is added until you say so.

  1. Read a card. Click it to apply it.
  2. Click × on a card you do not want, to dismiss it.
  3. When there is more than one card, click Apply all to apply them together, or tick the cards you want and click Apply selected.

Cards that add something to a desk also have a Choose where to add this button. Use it to put the result on this desk, on a new desk you name, or on another desk. See Review and apply Plexii's changes Coming soon.

If nothing in the recording sounds like a task, no suggestions appear.

Send the text to another widget

Use Send to… to put the text into a widget you can edit and keep working on.

  1. Choose the view you want to send: Summary, Cleaned, Speakers or Full. Send to… sends the text that is on screen.
  2. Click Send to… at the bottom of the note. A panel opens headed Where should this text go?
  3. To start a new widget, stay on New widget and click Sticky, Note, Markdown or Page. The new widget appears just below and to the right of the Voice note, titled Voice note summary, Voice note (cleaned), Voice note (diarised) or Voice note transcript, depending on the view you sent.
  4. To add to a widget you already have, click Append to existing instead. The number beside it says how many Sticky, Note and Markdown widgets are on the desk. Click one, and the text is added at the end of it, under a divider line with the date and time.

To close the panel without sending anything, click × or click outside it.

Record a new note in the same widget

The microphone button at the right of the top row (a camera for a video note) records a new note that replaces this one. Rest the pointer on it to see Record a new voice note (replaces this one).

  1. Click the microphone (or camera) button. Recording starts straight away, in the same mode as before.
  2. Speak, then click Stop.

The new recording and its text take the place of the old ones. To keep the old note, add another Voice note widget for the new recording instead. To switch between a voice note and a video note, add a new Voice note and choose the other button.

If something goes wrong

When the note cannot finish, it shows a message with up to two buttons:

  • Try again transcribes the saved recording again. It appears when the recording was saved, so you do not have to speak again.
  • Record new starts a fresh recording.

Common causes:

  • The microphone or camera is not allowed. On a Mac, open System Settings > Privacy & Security > Microphone (and Camera for video), turn PlexiDesk on, then click Record new. See Allow microphone, camera and screen recording on a Mac.
  • A key is missing for Cloud transcription. Add your OpenAI key in Settings > AI, or switch Voice · transcription provider to Local. Then click Try again.
  • Summary, Cleaned or Speakers will not load. These views need your own Anthropic key in Settings > AI. Without one, a new recording shows the transcript as it came back, and switching to one of these views shows a message instead. Add the key, click Try again, then choose the view again. Full works without it.

The title bar

Rest the pointer on the note to show its title bar. There you can rename the note, pin it to a corner of the screen, open it larger or in focus mode, and remove it. Right-click the title bar for the menu every widget shares. See The controls on every widget and Right-click menus.

Plexii and your voice notes

Plexii can read the text a Voice note is showing, so you can ask questions about what you recorded on a desk. A Living Doc or living page on the same desk includes it in its summary too. See Keep a living summary of a desk.

Good uses

  • Capture a thought before it goes. Add a Voice note to the desk you are working on, record a few sentences, and come back to the summary later.
  • Debrief after a call. Record what was agreed while it is fresh, read the Key points:, and apply any to-do list it suggests.
  • Talk a draft into shape. Explain the idea out loud, switch to Cleaned, then use Send to… > Page and edit it into a document.
  • Keep a running log. Keep one Note on a desk as a log, and use Append to existing to add each day's voice note to it, dated.
  • Show rather than tell. Record a short video note of yourself explaining a sketch or a screen you are holding up to the camera.

Common questions

Are the live captions the transcript? No. The captions are only a preview while you record and are not saved. The transcript is made from the recording after you press Stop.

Can I edit the transcript? Not in the Voice note itself. Use Send to… to put the text in a Sticky, Note, Markdown or Page widget, and edit it there.

Do I need an API key? For the transcript, only if you use Cloud transcription, which needs an OpenAI key. Local needs none. For the Summary, Cleaned and Speakers views and the suggested actions, you need your own Anthropic key. Full needs neither once the transcript is made.

Does Speakers recognise who is talking? No. It splits the conversation into turns from what was said, and labels them Speaker 1, Speaker 2 and so on. It suits a conversation between two people.

Can I play the recording back? Yes. A player sits under the text of every finished note, for both voice and video notes.

Will recording again keep my first note? No. The microphone button in a finished note replaces the recording and its text. To keep both, add a new Voice note for the second recording.

Still need help? Contact support at [email protected].