Putting a phone on the table is the easy part. The questions that actually decide whether you record are the ones most guides skip: what happens when the other person says no, who the recording belongs to afterwards, and whether the transcript will really tell you who said what.
Most guides to this end at “always get consent”. That is advice, not a mechanism, and it leaves the three questions a seller actually has unanswered. Here is how our recorder behaves in each case.
Pressing record does not start a recording. It moves the session into a consent step and stops there. The microphone is not opened, and no audio exists, until somebody confirms the other party agreed. The prompt you are shown is chosen by the consent policy rather than typed by you.
Declining does not pause or park the recording, because there is nothing to pause: the state machine moves the session to discarded and it ends there. No partial audio is kept, and there is no draft to find later. This is the question no other guide on this subject answers, and it is worth knowing the answer before you are sitting across a table.
When the people in a conversation resolve to different places, the gate applies the strictest mode that applies to any of them rather than the one most convenient to the recorder. When a jurisdiction cannot be resolved at all, the default treats it as the stricter case, so an unknown never quietly becomes permission.
Everything above describes how our software behaves. It is not legal advice, and it is not a statement about what the law requires of you. Whether you may record a given conversation depends on where the people in it are and on facts we cannot see, which makes it a question for your own counsel.
A recording you forget about is a liability rather than an asset. Capture stops automatically at four hours of elapsed time, so a session left running through an afternoon ends itself instead of quietly collecting a building's worth of conversation.
On stop, the metadata for the meeting is written first and the audio follows behind a storage seam, which is what lets the recording exist in one place and the record of it in another. Transcription runs from that stored audio rather than from the browser, so closing the tab does not strand the meeting half-processed.
Speaker labels are usually sold as a feature a tool either has or does not. That is not how it works. Attribution comes from the transcription engine that resolves for the recording: when that engine diarizes, the transcript is speaker-attributed, and when it does not, the transcript is accurate but flat. We would rather tell you which of those you are getting than print a tick in a comparison table.
The session is discarded. Because capture never began, there is no partial audio to delete and no draft left behind. The recorder moves from the consent step straight to discarded and the meeting proceeds unrecorded.
No. Pressing record moves the session to a consent step and stops. The microphone is not opened until somebody confirms the other party agreed, so the act of asking is not itself captured.
The gate applies the strictest mode that applies to anyone in the conversation rather than the most permissive one. If a jurisdiction cannot be resolved, the default treats it as the stricter case. Whether that satisfies your obligations is a question for your counsel, not for us.
Only when the transcription engine that resolves for that recording performs diarization. When it does, segments are speaker-attributed; when it does not, the transcript is flat. It is a property of the engine rather than a feature toggle, which is why we say which one you got rather than claiming it universally.
Yes. Capture stops automatically at four hours of elapsed time. No competitor page we read mentions any session limit at all, which is worth knowing if you have ever left a recorder running after a meeting ended.
No. Capture runs in the browser using the standard microphone and recording APIs, so the meeting is recorded from the device already in front of you. There is no bot to invite, which is the usual awkwardness when a tool designed for video calls is pointed at a table.