Frequently asked questions

You create an event in the web dashboard and upload your participant list, which automatically emails each person a one-tap consent + enrollment link. They install the CrowdScribe app on their phone, accept the consent disclosure, and spend about a minute recording a voice sample (and optionally a few face frames if you enable face capture for video-recording sessions). During the event, one or more designated phone(s) records audio in short chunks; CrowdScribe transcribes each chunk and labels every line with the speaker’s name. The host sees the transcript filling in live and can download a PDF or JSON copy once the event ends.

Free for up to 5 participants in events lasting an hour or less, up to 3 free events per month. Beyond that, pricing scales by attendee count and event length on a volume-discount curve — most small conferences land between $100 and $600 per day. See the pricing page for the live calculator. Events over 250 attendees are quoted individually but are extremely competitive.

Voice samples, face images, and the mathematical voiceprints derived from them are permanently destroyed within 72 hours of your event ending. We keep that 72-hour buffer so we can re-process the attribution if a customer flags a problem; once it expires, deletion is irreversible and automatic. Audio recordings get a 7-day retention; text transcripts are kept for 90 days, then deleted from our systems. Full schedule on the retention page.

No — face capture is off by default. Most events are audio-only and use only voice for speaker identification. You opt in to face capture at event creation if you want it, currently offered at no additional cost. Video recording is currently in beta and is only advisable for very large events where face capture on phones dedicated to recording (for example, trained on an audience microphone) is necessary for speaker identification due to poor audio quality or background noise.

Enrollment is optional for each participant. If someone skips it, they can still attend; their spoken contributions will be transcribed and labeled as “Unidentified speaker” in your transcript. As the host, you’re billed based on your guest list — whether someone enrolls or not doesn’t change the cost. To ensure an accurate transcription and enable a targeted name replacement for unidentified speakers, our system distinguishes between different unidentified speakers and the event cost is calculated based on the total number of unique speakers detected, whether they are enrolled or not.

Any iPhone running the CrowdScribe app. Available for Android soon. You designate one phone per session — typically a moderator’s — and give them a 6-digit code from the dashboard. The app records 15-second audio chunks and uploads them to our servers for transcription. The phone needs to stay connected to Wi-Fi or cellular for the duration of the session. Additional session recorders will be enabled in the near future.

Plug it back in and re-open the app. You can rotate to a different phone by issuing a new recorder code; the new phone picks up where the previous one left off in the same session. (While the app is designed to minimize battery drain, we strongly recommend plugging the recorder phone in for any session over 30 minutes.)

Very good for clear English in a typical conference setting — the underlying model is among the best automated transcribers available. Accuracy drops on overlapping speech, heavy accents, specialized vocabulary, and recordings made far from speakers (e.g., recorder phone left on a podium 40 feet from the audience Q&A). For best results, place the recorder phone within ~10 feet of active speakers.

Yes. Open the event in your dashboard and use the Danger Zone control at the bottom. You’ll be asked to type the event name to confirm. Deletion is immediate and irreversible — all participant biometric data, audio recordings, transcripts, and event metadata are destroyed on the spot. Manitou Research cannot recover any data after deletion.

Not yet. The mobile app checks your location at the start of enrollment and only proceeds if you’re physically in the U.S. This is a deliberate compliance posture — biometric data laws outside the U.S. (especially GDPR in Europe) are stricter than what we can support in our current beta. International rollout is on the roadmap. For now, all hosts and participants must be in the United States during enrollment and recording.

Yes. We execute DPAs and mutual NDAs for enterprise customers, security teams, and counsel running vendor diligence. A signed DPA discloses every subprocessor (including the specialized AI providers not named publicly), restates our retention and deletion commitments in DPA-standard form, and covers incident-notification timelines and audit rights. We accept the customer’s standard template if reasonable, or we’ll provide ours. Typical turnaround is under a week. Details and request instructions are on the enterprise legal page.