Voice and Speaker Separation Policy

• We do not identify or verify anyone by their voice, and we do not offer a feature that does.
• We do not create, collect or keep voiceprints, voice profiles or any other template of a person’s voice.
• We do not compare voices between recordings or between accounts to find the same person.
• We do not sell, lease or trade voice data, or disclose it except to the processor that transcribes it, as described below.

When speaker separation is used, our speech-recognition processor (Eleven Labs Inc.) analyses the audio of that one recording to tell the voices in it apart, and returns the transcript with anonymous labels — “Speaker 1”, “Speaker 2”. The labels mean something only inside that recording. If you rename a label, the name you type is stored as part of your content, encrypted on your device like the rest of it. With every request we tell the processor not to use its speaker-identification feature, which matches voices against previously enrolled speakers.

What we cannot promise. To tell voices apart, the processor may compute numerical representations of the voices while it handles the request. It has not given us a written commitment about whether it keeps those after the request is finished, so we do not claim that it does not. We do not receive, store or use any such representation.

In the United States, speaker separation is off unless you turn it on, and the app asks for your explicit consent before it does. Recordings can be transcribed without it. Because the voices in a recording belong to everyone who speaks, our Terms (Article 18) require you to tell the participants before you record that the recording will be transcribed and processed with AI, and to obtain the consent the law requires.

• Audio uploaded for transcription — the temporary copy on our servers is deleted as soon as transcription finishes; if that step is interrupted, an hourly cleanup removes any copy older than one hour.
• At the speech-recognition processor — once we have received and saved the transcript, we ask the processor to delete its transcription record for that recording. A copy may remain in its backups for up to about 30 days after deletion. The processor also states that it does not keep voice data more than three years after your last interaction with it, unless the law requires otherwise.
• Your recording and its speaker labels — kept in your library, encrypted, until you delete the recording (then 30 days in the trash) or your account (then a 7-day grace period). The recording is yours; we keep it only so that you can use it.
• When you delete your account — we delete your recordings from our systems after the grace period and ask the processor to delete any transcription record that remains. Our backups roll over after 7 days.

The processors that receive content, and what each keeps, are listed in the Processors table of our Privacy Policy.

Recordings are encrypted on your device before they are stored, and audio travels to the processor over an encrypted connection (TLS). The processor handles it in readable form while it transcribes — that is how transcription works — so it is not covered by storage encryption at that moment.

Write to hello@recnovo.com — including if you do not use Recnovo but appear in someone’s recording. How we handle access, deletion and appeals is set out in the United States section of our Privacy Policy. If we change this policy materially, we will tell you in the app or by email at least 7 days before the change takes effect.