AI notetaker for user researchers
In interview work the finding is often the exact sentence, not the gist of it. Wispr Notetaker records on your device, so nothing joins the interview, and its refined transcript names each speaker after the session. Each quote stays attributed to the participant who said it.
Why researchers need the transcript, not the summary
Most notetakers are built for people who want the gist. A researcher wants the opposite. The finding is often the phrasing itself: "I gave up and asked my manager to do it," said with that pause in the middle. Paraphrase it and you have lost the evidence.
Three things follow from that. First, the words have to be right, including the product names, feature names, and internal jargon your participant picked up from your own onboarding. Second, you need to know who said what, because a usability session with a participant, a stakeholder observer, and you is three voices with three different weights. Third, the record has to survive past this week, because synthesis happens after session nine, not session one.
Wispr Notetaker records on your device instead of sending a bot into the call, and gives you a live transcript during the session plus a refined transcript after it ends, with each speaker it can identify named. People should still know you are recording; capture without a bot does not change that.
A week in the life of a user researcher
The mix is lumpy. A discovery block of five 45-minute interviews in three days, then a usability round with a stakeholder sitting in, then a stakeholder readout where you present what you heard, then two weeks of tagging and writing while new sessions keep landing.
The friction shows up in the same places every round:
- A participant softens or performs once a named bot appears in the participant list
- The recruiter's calendar invite gets moved and the ad-hoc rescheduled call is the one you forget to record
- Back-to-back sessions blur, and the session that got saved turns out to be the wrong one
- Your product's own vocabulary comes back misspelled, so searching for it later returns nothing
- The quote you remember from session three takes 20 minutes of scrubbing to find
None of these are exotic. They are just expensive, because a lost session cannot be re-run with the same participant saying the same thing.
What Wispr Notetaker does in a research session
Nothing joins the call. Wispr Notetaker records on your device, which is why it works on Google Meet, Zoom, Microsoft Teams, Webex, and a Slack huddle without a separate integration per platform. A reminder to get consent stays on screen the whole time it records.
During the session you get a live transcript, plus a Thoughts tab for your own observations and a chat you can ask questions in. Speakers show as You and Them while it runs. After the session ends, the refined transcript arrives with names attached. You name anyone it could not identify, once.
Names and terms come from context rather than setup work. Wispr Notetaker reads the calendar invite for who is in the session, and it inherits your personal dictionary from Wispr Flow, so the product names and acronyms you already taught dictation are spelled right in your first interview.
What do you get out of each type of research session?
Each session type produces a different artifact: discovery interviews give you a verbatim transcript, usability tests with observers give you named speaker lines, and synthesis gives you answers with the source moment attached.
The last row is the one that changes a researcher's week most. You can ask across your meeting history in plain language and click through to the exact moment an answer came from, instead of reopening nine files. If your analysis happens in Claude or ChatGPT, you can connect Wispr Notetaker and query your transcripts, summaries, and notes there. Synthesis prompts then run on the real words instead of a pasted excerpt. That connection exposes Wispr Notetaker notes and briefs only; your dictations stay out of it.
How to run one session end to end
- Add your product's vocabulary to your personal dictionary in Wispr Flow before your first session, including plural forms and any acronym your participants say out loud.
- Keep the session on your calendar so the invite supplies participant names. Ad-hoc calls are detected too, with a one-tap start.
- Say out loud that you are recording, then start recording from the notification.
- Type your own observations in the Thoughts tab while the live transcript runs. Your typed notes and the summary stay separate, so editing one leaves the other alone.
- If you want a note in your own words without stopping capture, dictate it. It goes where you are typing and never lands in the meeting transcript.
- After the session, open the refined transcript and name any speaker still showing as Speaker 2. The name applies across the transcript, summary, previews, and chat.
- When you synthesize, ask across sessions by name and follow the citation back to the moment.
One thing to plan around: recording length is set to 2 hours by default, and you can raise it to 3 hours in settings before a long workshop.
Consent and privacy when your participants are strangers
Research consent is stricter than internal-meeting consent. You are recording someone who does not work with you, often under a signed form, sometimes on a topic they would not repeat in public.
Wispr Notetaker keeps a "get consent when transcribing others" notice in front of you the whole time it records. Transcript retention is yours to set, indefinitely by default or auto-delete after 1, 7, 30, 90, 180, or 365 days. Deleting a meeting is permanent across every signed-in device and takes the notes, summary, transcript, and speaker-identification data with it.
One thing to be precise about with your legal team: the SOC 2 Type II and ISO 27001 certifications cover Wispr Flow, on every plan. HIPAA-ready covers dictation with a signed BAA, and Wispr Notetaker is not available on a BAA account. If your studies touch protected health information, that is a reason to wait.
Keep your history when you switch
Nobody wants to strand two years of interviews. Coming from Granola or Otter, one-click import brings your history into Wispr Notetaker. Coming from anything else, the archive stays where it is and stays searchable: connect your old notetaker and Wispr Notetaker to Claude, and one chat can search both. Name both tools in the prompt when you want the old sessions, because AI tools tend to default to one source.
You do not have to cut over mid-study either. Run both tools on one session and compare the transcripts on the thing you actually care about, which is whether your participant's exact sentence survived and whether the right person is credited with it.
What import does not carry: your tags, your codebook, and your analysis structure live in your research tooling, not in a notetaker, and they stay there. Sharing is per-note, to named people, so there are no shared folders or project spaces to rebuild either.
The one session that tells you
Pick your least precious upcoming interview, ideally one with three people on the call and at least one product name your current tool has been mangling. Add that word to your personal dictionary first. Run the session, then read the refined transcript against your memory of it.
You are checking two things. Did the sentence you would put in the readout come back exactly as it was said? And is it attributed to the participant rather than to the stakeholder who agreed with them a second later? If both are right, the rest of your analysis is built on something solid. If either is wrong, you would have found out in the readout instead, in front of the people who act on your work.
Record your next research session and read the transcript before you decide.
Frequently asked questions
Can I get a verbatim transcript instead of a summary?
Yes. The live transcript runs during the session and a refined transcript arrives after it ends, with names attached. The summary is built from that refined transcript and kept separate from it, so editing or deleting one leaves the other alone and your own edits are never silently overwritten.
How does it handle a session with a participant and two observers?
You and the two observers show as You and Them while the session runs; names arrive with the refined transcript. Anyone the invite did not cover you name once. Speaker assignment works only on meetings you own, and a voice heard on two channels at once stays a plain Speaker label rather than being guessed.
Will it spell my product's feature names correctly?
It inherits your personal dictionary from Wispr Flow, the same list dictation uses for names, jargon, and acronyms. Terms you already taught Flow show up spelled right in your first interview. Add both singular and plural forms, and keep entries short.
How do I find a quote from a session three weeks ago?
Ask in plain language across your meeting history, and every answer links back to the source moment you can click into. You can also search past notes by meeting name and attendee, which works when the participant's name is on the invite.
Can I export a transcript to my analysis tool?
Sharing is per-note rather than a bulk export. A share link shows the summary only; the full transcript needs an email invite, and invited people get viewer access, not editing. If you work in Claude or ChatGPT, you can connect Wispr Notetaker and query the transcripts there instead of moving files around.
Do participants know I am recording?
They know because you tell them. Nothing appears in the participant list, so the disclosure is yours to make out loud or on the consent form. In the app, Wispr Notetaker keeps a "get consent when transcribing others" notice in front of you the whole time it records, and your organization can require you to confirm consent before recording starts.
Can I delete a participant's transcript on request?
Yes. You can delete any meeting, transcript, or note at any time, and deletion is permanent across every signed-in device, taking the notes, summary, transcript, and speaker-identification data with it. You can also set transcripts to auto-delete after 1, 7, 30, 90, 180, or 365 days.
