Meeting notes for user researchers
Listenr records research interviews from your Mac with no bot in the session and transcribes them on-device. The transcript is an editable document with speakers, so the quote you put in the report is the one the participant said.
The transcript is the deliverable
Read and correct it line by line, with speakers on every line, then export the whole session as Markdown.
Speakers, separated on your Mac
You are “You”, everyone else is Speaker 1, 2. Names are added on Google Meet, Zoom desktop and Microsoft Teams when the evidence is clear.
Consent-form facts
Audio is transcribed on your Mac, no voiceprints are stored, and recordings are not kept unless you ask.
Interviews in other languages
Parakeet Ultra covers 25 European languages with detection. SenseVoice covers 50+ including Chinese, Japanese and Korean.
Listenr is an AI notetaker for user research interviews on Mac. It records the session from your Mac with nothing joining the call, transcribes on-device by default, and treats the transcript as an editable document with speakers, not just raw material for a summary. That matters in research, where the exact wording is often the finding.
Quotes you can put in a report
Every line shows who said it and when. You can correct a line where the transcription misheard a product name, and export the whole session as Markdown into your analysis tool. If the language was wrong, retranscribe the session, or just one section, in the right language.
How speakers are attributed
Knowing where the speaker label comes from tells you how far to trust it.
- You: your side comes from your microphone, so the moderator is always separate from the participants.
- Everyone else: voices are separated on your Mac when the session ends and labelled Speaker 1, Speaker 2. The default engine computes no voice embeddings, so no voiceprint is created or stored.
- Names on Google Meet: with the Listenr browser extension (Chrome, Brave, Edge, Arc or Safari), Listenr reads Meet’s participant list, mute states and speaking indicators. It does not read audio, video or captions.
- Names on Zoom desktop: after you switch on the Zoom Accessibility reader, Listenr reads Zoom’s participant panel the same way. It never clicks or types. Zoom in a browser is not covered.
- Names on Microsoft Teams: Listenr names speakers in Teams sessions too.
- Other apps: voices stay Speaker 1, Speaker 2.
A name is assigned only when the evidence is clear, for example when the participant list shows only one remote person was unmuted. Otherwise the line keeps its Speaker number. A one-to-one interview is the easiest case. In a focus group more lines stay unnamed, which can mean resolving them by hand.
An optional setting, off by default, lets a local model also pick up names from what people say (“thanks, Anna”).
What you can accurately put in a consent form
- Audio: transcribed on your Mac with every engine except OpenAI cloud transcription, which you would have to select.
- Recordings: not kept by default. A temporary recovery copy is deleted once transcription succeeds, kept only if transcription runs into trouble, and purged after 14 days at most. If you choose to keep a recording, it stays on your disk.
- Voiceprints: none stored.
- Notes: a local model or Apple Intelligence sends nothing off the Mac. Your own Anthropic or OpenAI key sends the prompt to that provider. Listenr Cloud sends it through Listenr’s servers.
- Telemetry: none in the app.
The notes prompt includes participant display names and the meeting chat, but no email addresses. If participants join under their real names, those names reach the notes model you choose. For a strictly local study, use a local notes model. You can also switch off the Meet extension or Zoom reader, and chat capture, in settings.
Nothing announces the recording. No bot appears in the participant list and no banner is shown, so a participant knows they are being recorded because you told them.
Sharing sessions with your team
Notes deliver to Obsidian or Apple Notes as local files, or to Notion on Pro, and any session exports as Markdown. If participants were recruited under a pseudonym, a Markdown export is the safer way to share: a share link (Pro) creates a public page with the display names, notes and transcript. What leaves your Mac explains how share links work.
Finding the same theme across sessions
Every session becomes searchable text on your disk. Search it in the app or with the listenr command line, or ask Claude Desktop or Claude Code, through the local MCP server, which participants raised setup cost. Both are free on every plan and read only the database on your Mac.
When to pick something else
Listenr records only the Mac in front of you, so it cannot capture a session a colleague moderates. A notetaker bot can, and it shows participants a visible recorder, which some studies want in the record. If most of your sessions run in Zoom in a browser or another app without names, expect Speaker numbers. The notetaker bots comparison covers the alternative.
Common questions
What is a good AI notetaker for user research interviews?
For interviews, the transcript matters more than the summary, and a bot in the call changes the session. Listenr records from your Mac with nothing joining, transcribes on-device, and gives you an editable transcript with speakers that exports as Markdown. It runs on macOS 15 or later with Apple Silicon.
How does Listenr know which participant said which line?
Your side comes from your microphone and is labelled You. Other voices are separated on your Mac after the session and labelled Speaker 1, Speaker 2. On Google Meet (with the Listenr browser extension), the Zoom desktop app (with its Accessibility reader switched on) and Microsoft Teams, Listenr adds a name only when the evidence clearly points to one person; otherwise the line keeps its Speaker number.
Does Listenr create voiceprints of participants?
No. The default speaker engine separates voices without computing voice embeddings. The alternative engine, used for very large meetings or as a fallback, computes temporary embeddings that are discarded after the pass. Nothing is stored or compared across meetings.
What can I tell a participant about where the recording goes?
The audio is transcribed on your Mac unless you choose OpenAI cloud transcription. By default no recording is kept: a temporary recovery copy is deleted once transcription succeeds, and any left over is purged after 14 days. What leaves the Mac for the notes depends on the notes model you choose.
Does Listenr tell participants they are being recorded?
No. Nothing joins the call and no banner appears, and the Zoom reader and Meet extension only read what your own meeting app already shows you. Disclosure is your responsibility. Consent rules differ by country and state, and some require everyone on the call to agree.
Can I transcribe interviews in other languages?
Yes, by picking an engine. The default, Apple’s on-device speech recognition, uses one fixed language. Parakeet Ultra covers 25 European languages and detects which is spoken, SenseVoice covers more than 50 including Chinese, Japanese and Korean, and all of them run on your Mac. You can also retranscribe a session or a section in a different language.
Install it before your next call.
Recording, transcription and notes from a model on your Mac are free. macOS 15 or later, Apple Silicon.