How does the AI call recorder work?

The recorder captures a sales call in your browser, transcribes it, and writes back a summary, the objections you heard, a read of the room, and the next steps. You can turn those next steps into real tasks with one click. It's included on every paid plan — Solo, Pro, and Enterprise.

Start one from any contact, company, or deal page, from a meeting row on your dashboard, from the Calls page in the sidebar, or from the command palette. Already have the audio? Upload it instead.

What gets captured

Two modes, picked in the consent gate before anything starts:

• My microphone only — just your side. The transcript comes back as one unlabelled block.
• My mic + the other side — two tracks, merged with timestamps, so the transcript reads as Rep: and Them:. This is the one worth using.

Capturing the other side uses your browser's screen-share picker, which means:

• It works in Chrome and Edge on desktop only. Firefox and Safari don't offer tab audio.
• On a phone, neither mode captures a phone call. iOS exposes call audio to nothing, so if you're calling from your mobile the closest equivalent is a voice note recorded straight after you hang up — see calling a list from your phone.
• For a call in a browser tab (Google Meet, a dialer, a web client), pick the tab and tick Share audio.
• The Zoom and Teams desktop apps are not browser tabs. Pick Entire Screen and tick Share system audio. That option exists on Windows and not on macOS, so on a Mac you'll get your own side only.

The recorder throws away the video track immediately. Only audio is ever read.

What we keep, and what we don't

We never store the audio. It is not written to disk and not written to a database column. It lives in the server's memory only until transcription finishes, then it's gone. If the server restarts mid-transcription, that recording is lost and cannot be recovered, because there is nothing left to retry from. Upload the file again if you still have it.

We do keep, until you delete the recording:

• the transcript, with its timestamps and speaker labels
• the summary, objections, sentiment, buying signals, risks and next steps
• the consent line you read out
• any tasks and activity entries created from the call

Deleting is permanent. It removes the transcript and summary outright, and unlinks the recording from the meeting it was attached to. There is no trash and no undo. Tasks you already created from the next steps are not deleted, because by then they're your tasks.

Consent is your responsibility

The consent gate shows a line to read out loud and asks you to confirm you did. We store our own copy of that line against the recording, so the text you read and the text we log are always the same.

That's a sensible default, not legal advice, and not a compliance guarantee. Recording law varies: some places need only your consent, others need everyone's, and some require specific wording. You are responsible for knowing the rule where you and the other person are sitting. DealArena doesn't check your jurisdiction and doesn't block a recording based on it.

If someone says no, don't record. Take notes the old way.

Your daily limits

Two rolling 24-hour caps, per user:

• 20 recordings
• 5 hours of total audio

Both are rolling windows, not calendar days, so the oldest recording ages out 24 hours after you made it. You can see where you stand at Settings → Call recorder and in the header of the Calls page. Hit either cap and new recordings are refused until the window rolls.

There's also a separate 20 uploads per hour limit applied per network address. You'll only meet it if you're backfilling a stack of old recordings in one sitting, and it clears on its own within the hour. If you share an office network, you share that limit.

Recording draws on the same monthly usage as everything else. A recorded minute costs 1 unit per audio track, so a two-sided call is 2 units a minute and a half-hour call is 60 units. You are charged on the length the upload claims, bounded by what the file size proves it can be, and that charge is not revised afterwards. A daily cap (20 recordings or 5 hours in any 24 hours) still applies as an abuse brake. See how usage works.

Uploading a recording you already have

Pick Upload a file I already have in the recorder, choose the file, and it goes through the same transcription and summary as a live recording. Accepted formats: webm, ogg, m4a, mp3, wav, flac.

The consent gate still applies. A call recorded somewhere else still needed everyone's agreement at the time.

Uploaded files are single-track, so you won't get Rep: and Them: labels — the transcript comes back as one block. That's a limit of the file, not of the transcription.

Why won't my file upload?

The limit is 20 MB per track, and it's the size that usually bites. Here's roughly what one hour of audio weighs:

Format	60 minutes
WAV, 44.1 kHz stereo	~605 MB
WAV, 16 kHz mono	~110 MB
MP3 or AAC, 128 kbps	~55 MB
MP3 or AAC, 64 kbps mono	~28 MB
Opus, 32 kbps mono	~14 MB

A recording DealArena made is Opus at 32 kbps mono, so it re-uploads without trouble. A raw export from Zoom or a phone usually won't fit.

To get under the limit:

• Export audio only, not video. Most call platforms have this option, and it's the single biggest saving.
• Export mono, not stereo. That halves it.
• Re-encode at 32 kbps or lower. Speech stays perfectly intelligible; Whisper doesn't need music quality.
• Split a long call in half and upload each part as its own recording.

Other rejections you might see:

• "That file type isn't supported" — we check the file's actual contents, not its name. Renaming a .mov to .mp3 won't get it through. Re-export it properly.
• "You've hit today's recording limit" — one of the two daily caps above.
• "Transcription isn't configured on this environment" — a setup problem on our side, not yours. Check Settings → Call recorder; it says plainly whether transcription is available.

Linking a call to the right person

A recording started from a contact page is attached to that person and their company automatically. One started from a company or deal page is attached there, and you can add the person afterwards.

If the recording lost its link (the prospect was deleted mid-call, or you started it from the command palette with nothing selected), it's saved rather than thrown away, and it shows up under unlinked only on the Calls page. Open it and set the contact from the detail page.

When something goes wrong

• Partial — the transcript came through but the summary didn't. The transcript is the expensive part and it's safe. Hit Retry summary.
• Failed — transcription itself didn't complete. The recording page tells you why.
• The tab audio was silent — you almost certainly shared the tab without ticking Share audio, or shared the wrong tab. The recording still has your side.

Closing the tab, navigating away, or stopping the share all end the recording cleanly and upload what was captured. You won't lose a call by clicking the wrong thing. Recordings stop automatically at 75 minutes, with a warning at 65.