COMPARED
Where your audio actually goes
Every scribe here is good at something. They differ on one thing that no settings page can undo: whether the recording leaves your machine. That is what this table sorts by. Prices and details checked September 2026, with each vendor's own page linked on its comparison.
| Tool | Where the audio goes | Processing | Bot in the call | Open source |
|---|---|---|---|---|
| Memory Notch | Nowhere | On-device, on the Neural Engine | No | No |
| Otter.ai | To Otter's servers | Cloud | Yes, a participant joins the meeting | No |
| Fireflies.ai | To Fireflies' servers | Cloud | Yes, a participant joins the meeting | No |
| Granola | To a cloud transcription provider, then deleted | Cloud, with local capture | No | No |
| Fathom | To Fathom's servers | Cloud | Yes, a participant joins the meeting | No |
| Abridge, Nuance DAX and the clinical scribes | To the vendor's HIPAA-covered cloud | Cloud | No, it listens in the room | No |
| Hyprnote | Nowhere | On-device | No | Yes |
| Meetily | Nowhere | On-device | No | Yes |
| MacWhisper | Nowhere | On-device | No | No |
Bot-free is not the same as private
The marketing around notetakers has settled on "no bot in your meeting" as the privacy feature. It is a real improvement to the room: nobody sees a robot join, and people talk normally. It is not a statement about your data.
Granola is the clearest case, and to its credit the company documents it plainly. The app captures microphone and system audio on your Mac with no bot at all, then passes that audio to a cloud transcription provider while the meeting is happening. The audio is deleted once the transcript exists; the transcript and notes are kept on AWS in the United States. That is a thoughtful design. It is still a design in which your conversation is transmitted.
The distinction worth holding on to: who is in the room and where the words are processed are two different questions, and only the second one survives a subpoena, a breach or a change of ownership.
When a cloud scribe is the right answer
Often. If you need a shared team archive, CRM sync, conversation analytics across dozens of reps, or a clinical note written into an EHR with a signed BAA behind it, the cloud tools do things an on-device app on one laptop cannot. Fireflies and Otter are built for teams. Abridge and Nuance DAX are built for health systems and are genuinely good at clinical language. Fathom has the most generous free tier of the group.
On-device is the right answer when the conversation is the sensitive thing and you are the only one who needs the notes.
The comparisons
- Memory Notch vs Otter.aiThe best-known meeting transcriber, with a bot that joins the call.
- Memory Notch vs Fireflies.aiA meeting recorder built around the CRM, aimed at revenue teams.
- Memory Notch vs GranolaA Mac notepad that captures the call without a bot, then transcribes it in the cloud.
- Memory Notch vs FathomA generous free tier for recorded calls, with a bot on the invite.
- Memory Notch vs Abridge, Nuance DAX and the clinical scribesAmbient documentation built for clinicians, sold to health systems.
- Memory Notch vs HyprnoteAn open-source, local-first AI notepad. The closest thing to this app in spirit.
- Memory Notch vs MeetilySelf-hosted meeting notes in Rust, with local diarization and Ollama summaries.
- Memory Notch vs MacWhisperOn-device Whisper transcription for Mac, aimed at files rather than live calls.