1. Choose the artifact before you choose a microphone
An editable chat message, an audio clip, captions, a meeting transcript, and a Copilot prompt have different recipients, storage behavior, permissions, and review points. The word “microphone” does not make them interchangeable.
Start with the result you need. If colleagues should receive normal text in a chat or channel, dictate into the editable compose box. If they should receive your voice, send a clip. If the speech belongs to a meeting, choose captions or transcription deliberately.
2. Dictate an ordinary Teams chat message on a Mac
Microsoft’s current Teams chat guide says to compose the message in the box at the bottom of the chat, then select Send or press Enter. It does not document a normal-chat Dictate button in that workflow.
Click inside the compose box, confirm the text cursor is active, and trigger the shortcut shown under System Settings → Keyboard → Dictation. Speak, stop Dictation, edit the visible text, and send only after names, dates, links, amounts, recipients, and attachments are correct.
- Open the intended chat, channel, or meeting-chat composer.
- Place the cursor in the editable compose box.
- Dictate, stop, review the visible message, then send.
3. Keep dictation and the Send action separate
Microsoft documents Enter as sending a normal chat message. The expanded Format editor provides a safer surface for longer text, paragraphs, lists, and links.
Do not use a spoken “new line” or an uncertain keypress as a substitute for reviewing the composer. Stop listening first. Check the destination and the full text, then make the final Send action explicit.
4. Use an audio clip only when you intend to send audio
Teams can record up to one minute of audio and send it to a one-on-one chat, group chat, or channel discussion. Microsoft places the workflow under Actions and apps → Record a video clip, where Camera can be switched to audio-only recording.
The clip is added to the message and sent into the conversation. Microsoft says shared clips have auto-generated captions and can expose a downloadable transcript. That is a retained communication artifact, not a temporary way to produce an ordinary editable message.
5. Separate live captions from a saved meeting transcript
Live captions display what Teams detects during a meeting. Microsoft says captions are not saved and caption data is permanently deleted after the meeting. They are for real-time access, not a durable meeting record.
Live transcription captures speech with speaker names and timestamps. Microsoft says the transcript becomes available through the meeting chat and Recap, is stored in the organizer’s OneDrive for Business, and remains accessible until an organizer deletes it. Recording and transcription permissions can be controlled by meeting policy and role.
6. Do not transfer Copilot Chat dictation to ordinary Teams chat
Microsoft 365 Copilot Chat has a documented Start dictation control in its prompt box for signed-in work or school accounts. Microsoft says it transcribes the spoken prompt in real time and tells users to review the result.
That control belongs to Microsoft 365 Copilot Chat. Its existence does not establish a matching dictation button in the ordinary Teams message composer.
7. Map privacy and retention to the exact path
For macOS Dictation, Apple tells users to inspect the disclosure under System Settings → Keyboard → Dictation to see whether general voice inputs and transcripts are processed on-device for the current Mac, language, and settings. The resulting message still becomes part of the Teams conversation and workspace controls.
A Teams audio clip sends recorded media into the conversation. Live captions use Microsoft automatic speech recognition but are not saved. Meeting transcripts are stored in the organizer’s OneDrive until deleted. Copilot Chat dictation uses its own Microsoft voice workflow. Review the policy for the path you actually activate.
8. Add IraVoice only for the distinct local cross-app need
IraVoice 0.7.1 is an option for macOS 26+ Apple-silicon Macs when the same held-key workflow should produce editable text across supported chat, email, document, and developer fields. Speech recognition and supported formatting run on the Mac after model setup; paid continuation needs one license-key check.
IraVoice does not record Teams meetings, create meeting transcripts, caption calls, send audio clips, join a Teams workspace, or press Send. It targets the focused editable field exposed through macOS Accessibility and falls back to the clipboard when direct insertion is uncertain.