Where Granola Keeps a Transcript and How to Export It

Granola does not keep audio or video recordings of your meetings. On desktop, Granola transcribes in real time, caches the audio only until transcription is done, then deletes it.

At Layer3Labs, we build workflow automations for small and mid-sized teams, and what a note-taker stores sets what those automations can reach.

Granola Transcript Generation and Audio Retention

Granola does not save or retain audio or video files from any meeting. According to Granola's security documentation, the desktop software on macOS and Windows converts audio into text in real time. Audio is cached during the meeting for transcription only. Once transcription is complete, Granola deletes it from its systems and any third-party services.

To get the text out, click Copy all in a note's transcript panel, or request a comma-separated values (CSV) file of every note and transcript from Settings → Profile → Generate CSV. Granola emails the CSV within a few hours. The mobile apps are built for in-person meetings.

Mobile note-taking follows a slightly different sequence with the same retention outcome. On iOS and Android, Granola transcribes after the meeting from temporarily cached audio. In Granola's feature requests page, Granola confirms it has no plans to support video or audio recording playback.

Granola uses transcription providers such as Deepgram and Assembly, and AI providers such as OpenAI and Anthropic. Granola says it does not allow third parties such as OpenAI or Anthropic to train their models on your data.

• Desktop transcription: Processes audio in real time and deletes the cache immediately upon call completion.
• Mobile transcription: Caches audio locally during the conversation, transcribes immediately after the call, and deletes the cache.
• Audio playback: none; Granola keeps no recording.
• Processing partners: Deepgram and Assembly handle speech-to-text; Granola does not allow third parties to train on your data.

Desktop and Mobile Transcription Mechanics

Granola requires manual activation and captures combined system audio alongside your microphone input. Granola never begins transcribing automatically without a user explicitly clicking into a note, as outlined in Granola's transcription guide. There is no virtual bot joining the meeting room or appearing in the attendee list.

Desktop capture works with Google Meet, Zoom, Microsoft Teams, Slack Huddles, and internet calling software by listening to the computer's system audio output mixed with the primary microphone. Transcription stops automatically under three specific conditions: when call termination is detected, when fifteen minutes pass without incoming audio, or when the host computer enters sleep mode.

Speaker identification defaults to labels for Me and Them. On macOS and Windows, Granola provides enhanced speaker tags for Google Meet, Zoom, and Microsoft Teams. Mobile transcripts assign labels as Speaker A and Speaker B. Granola supports 31 languages on desktop and 17 on mobile, allowing desktop meetings to transcribe multiple spoken languages within a single session. Teams can also define personal jargon and workspace jargon in settings to improve recognition of technical terms.

• Manual start: Nothing is transcribed until you open a note.
• Combined audio stream: Granola reads all computer sound together and cannot isolate audio from individual applications.
• Automatic shutoff: Capture stops at call end, after 15 minutes of silence, or when the computer sleeps.
• Speaker labels: Desktop uses Me and Them or platform tags, while mobile assigns Speaker A and Speaker B.

Methods for Viewing and Copying a Granola Transcript

On desktop, click the waveform icon in a note to open the transcript panel. Click Copy all to copy the whole transcript.

In the transcript panel you can search the text, copy single chunks and delete chunks. You cannot edit transcript text, so a misheard word stays unless you delete that whole chunk.

On iOS or Android, tap the three-dot menu on a note and choose View transcript. From there you can read the transcript and share it.

Note and Transcript Export Options

Clicking Copy notes inside an open meeting copies only your AI-enhanced notes in markdown format, leaving out the underlying speech transcript. Send notes via email opens a Gmail or Outlook draft of your notes.

To export complete transcripts, navigate to Settings, choose Profile, and click Generate CSV. According to Granola's export documentation, Granola generates the file and emails you a link within a few hours. The CSV includes each note's title, summary, transcript and other basic details.

Bulk CSV exports include all historical meetings, including notes older than 30 days on the free Basic tier. The download link remains active for 24 hours, and accounts are limited to one export request per 24 hours. On Enterprise plans, data exports are disabled by default until an administrator enables the feature. Granola does not offer an organization-wide administrator export, and public web links shared from Granola show summaries while hiding raw transcripts.

• Markdown copy: Copies only the AI-enhanced notes, not the transcript.
• Bulk CSV export: Delivers titles, summaries, and complete transcripts via an emailed link within hours.
• Historical access: includes notes older than 30 days, even on Basic.
• Shared link privacy: Public web view links display AI summaries but never expose the raw transcript.

Storage Locations and Retention Policies

All Granola meeting notes and transcripts reside on Amazon Web Services (AWS) servers located within the United States. Granola holds SOC 2 Type II certification and encrypts customer data at rest and during transmission using TLS. Granola keeps notes and transcripts until you or an admin set a deletion policy.

Individual users on any plan can establish a personal auto-deletion rule under Granola's transcript auto-deletion settings. Available schedules include 1 day, 1 week, 1 month, 3 months, 6 months, 1 year, or Off. When an individual activates or modifies an auto-deletion schedule, Granola enforces a mandatory one-week cooldown period before deletions begin. Deleting transcripts removes only the verbatim text; your structured notes and summaries remain intact.

Enterprise plans support workspace-wide transcript retention policies managed through Granola support, which take effect immediately without a cooldown window. On Basic and Business, Granola may use anonymised data to improve its own models by default; turn off 'Use my data to improve models for everyone' to opt out. Enterprise workspaces have model training disabled by default across all team members.

Limitations of the Granola Transcript Architecture

Granola does not accept pre-recorded audio file uploads and cannot isolate audio streams from separate desktop applications. Users cannot import MP3, WAV, or M4A files from external dictaphones or older call archives into Granola for transcription.

Because Granola captures desktop system audio, any background sound playing through your computer will merge into the recording stream. A video playing in a browser tab or an alert sound from another app during a meeting will enter the transcription model alongside attendee voices. Mute other apps and close media tabs before you start the note.

The absence of an audio file means that once a meeting concludes, there is no way to retrieve an audio snippet for legal verification or audio production. Furthermore, the web interface at notes.granola.ai functions solely for reviewing and editing existing notes, meaning users cannot capture or transcribe live calls through a browser.

• No audio imports: Pre-recorded audio files cannot be uploaded for transcription.
• Mixed audio capture: System sounds and media playback merge directly into the transcript.
• No audio recovery: Deleted audio caches cannot be retrieved under any circumstance.
• Web application limits: Browser access supports viewing notes but cannot capture live speech.

Tradeoffs Between Audio Recordings and Text Transcripts

Granola keeps no audio, so you cannot replay a call to check tone or exact wording. Sales coaching and dispute reviews can need that playback. For those jobs, use a note-taker that keeps the recording; our Granola vs Fathom comparison covers that trade-off.

If you handle sensitive client information, no stored audio means one less copy of a conversation to secure; our AI note-taker consent guide covers consent. As of October 2026, Granola's pricing plans start at $0 for Basic, $14 per user monthly for Business, and $35 per user monthly for Enterprise.

Granola is not suitable for organizations that require legal audio evidence or verbatim voice verification. Our assessment would change if Granola introduced optional encrypted audio archiving or pre-recorded audio file imports. If you already use Granola, open Settings → Profile and click Generate CSV to see exactly what your transcripts contain.