Tome is a macOS application designed to seamlessly capture meetings and voice memos while ensuring complete data privacy. By utilizing advanced transcription technology with the Parakeet-TDT v3 engine, Tome transcribes speech locally and structures the output into Markdown files that fit directly into an Obsidian vault, eliminating the reliance on cloud services or API keys.

Key Features

• Local Transcription: Conduct transcription on your own device using Parakeet-TDT v3, ensuring that all captured data remains secure and private.
• Vault-Native Output: Generate structured .md files complete with YAML frontmatter, enabling direct integration with your Obsidian vault without the need for proprietary exports or manual formatting.
• Multilingual Support: Efficiently capture dialogues in 25 European languages with automatic language detection, all while maintaining data privacy as no information is transferred online.
• Call Capture: Filter and record audio specifically from conferencing applications like Teams, Zoom, and Slack, removing extraneous sounds to deliver clear transcripts.
• Voice Memo Functionality: Quickly jot down thoughts or notes with a dedicated mic-only mode, allowing for easy organization while keeping meeting transcripts uncluttered.
• Speaker Diarization: Automatically differentiate between speakers after the session through advanced audio processing, providing clarity in shared discussions.
• Automatic Silence Detection: Enable the application to automatically stop recording when 120 seconds of silence is detected, preserving storage and focusing on relevant content.
• User Privacy Focus: No storage or network activity is involved in the transcription process; only text transcripts are saved, ensuring that user conversations remain confidential.

Operational Workflow

The workflow is straightforward:

speak → capture → vault → agent → knowledge base

1. Capture: Tome captures audio from both microphone and system outputs.
2. Transcribe: It utilizes voice activity detection (VAD) to recognize speech, providing efficient and accurate transcription.
3. Diarize: After the session, the system separates audio streams by speaker for improved readability.
4. Output: Structured Markdown files are created with essential metadata and saved to the specified Obsidian vault directory.
5. Agent Processing: An AI agent, if deployed, can further process the structured notes for tasks, follow-ups, or knowledge accumulation.

User-Friendly Design

Tome focuses on a user experience that integrates seamlessly into the workflow of consultants, researchers, or anyone who conducts meetings frequently. By simplifying the capture and transcription process into a single, efficient tool, Tome addresses the common pain points of note-taking during conversations and enhances productivity without compromising on data privacy.