Question 1

What is Parrot Scribe?

Accepted Answer

Parrot Scribe is a professional-grade local transcription tool for macOS. I built it using WhisperKit and optimized it for the Apple Neural Engine to provide 100 percent on-device processing.

Question 2

Is Parrot Scribe private?

Accepted Answer

Yes. Privacy is my core standard. All audio and text stay on your Mac. I never see your data because I designed the app to process everything locally.

Question 3

Does it require a subscription?

Accepted Answer

No. I believe you should own your tools. Parrot Scribe is a one-time purchase with unlimited transcription, no per-minute fees, and no usage caps.

Question 4

What are the system requirements?

Accepted Answer

Parrot Scribe requires an Apple Silicon Mac (M1 or later) running macOS 15 Sequoia or newer.

Question 5

How does it compare to other tools?

Accepted Answer

Unlike cloud-based tools (Otter, Fireflies), Parrot Scribe is 100% private. Compared to other local tools, I offer a native macOS 15 experience with AI-native TOON output and MCP integration for AI agents.

Question 6

What is the TOON format?

Accepted Answer

TOON (Token-Oriented Object Notation) is a compact, human-readable format designed for LLMs. Maximum context, minimum tokens. You can learn more about the standard at github.com/toon-format/toon.

Question 7

What is the MCP server?

Accepted Answer

The MCP (Model Context Protocol) server is built into Parrot Scribe. It lets tools like Claude Desktop and Cursor access your transcripts - both real-time and historical. It uses a Unix domain socket with a signed stdio shim for secure communication, features 3-tier permissions, and is disabled by default for zero attack surface when not in use.

Question 8

How do I move my data to a new Mac?

Accepted Answer

Encryption keys are bound to each Mac's Secure Enclave and cannot be transferred. Your transcripts are device-bound by design. Your license can be deactivated on your old Mac and activated on the new one.

Question 9

What is the difference between Remove and Forget for speakers?

Accepted Answer

Remove deletes a speaker's voice model but preserves their sessions - you can re-enroll them later from session embeddings. Forget is a GDPR deletion that permanently removes the voice model AND all sessions containing that speaker. Forget is irreversible and may affect sessions shared with other speakers.

Downloads & Release Notes

Version Catalog