SamScribe - AI-Powered Local Transcription App

January 10, 2026 · View on GitHub

An open-source macOS app that captures and transcribes audio from your microphone and meeting applications (Zoom, Teams, Chrome) in real-time, with cross-session speaker recognition.

Uses voice embeddings to remember speakers across sessions - name someone "Bob" once, and the app recognizes them automatically in future recordings.

Built with Swift, SwiftUI, and FluidAudio. All data stays local on your Mac.

Demo

Features

  • 🎤 Real-time Transcription: Transcribe audio as you speak with minimal latency
  • 👥 Speaker Diarization: Automatically identify and separate different speakers
  • 🔄 Cross-Session Recognition: Speakers persist across all recordings with voice embeddings
  • 📝 Editable Transcripts: Click any segment to edit transcription text inline
  • ✏️ Speaker Management: Rename speakers (e.g., "Speaker 1" → "Bob") with pencil icon
  • 💾 Persistent Storage: All recordings saved locally with SwiftData
  • 🔍 Per-Process Capture: Isolate audio from specific applications (Zoom, Teams, Chrome)

Use Cases

  • Meeting Notes: Automatically transcribe Zoom/Teams meetings with speaker identification
  • Interviews: Record and transcribe interviews with automatic speaker separation
  • Podcasts: Transcribe podcast recordings with persistent guest recognition across episodes
  • Lectures: Capture classroom discussions with speaker attribution
  • Research: Qualitative research interviews with cross-session participant tracking
  • Accessibility: Real-time captions for audio content

System Requirements

  • OS: macOS 14.0+ or newer
  • XCode: Version 16.0 or newer
  • Permissions: Microphone access and Screen Recording access

Development Installation

  1. Clone the repository:

    git clone https://github.com/Steven-Weng/SamScribe
cd SamScribe

  2. Open in Xcode:

    open SamScribe.xcodeproj

Usage

Basic Usage

  1. Start Recording: Click "Start Transcribing" button in the toolbar
  2. Grant Permissions: Allow microphone and screen recording access when prompted
  3. Speak or Join Meeting: Audio is captured and transcribed in real-time
  4. View Transcripts: Transcription appears in the main detail view as you speak
  5. Edit Speakers: Click the pencil icon next to any speaker name to rename them
  6. Edit Text: Click any transcript segment to edit the text inline
  7. Stop Recording: Click "Stop Transcribing" when finished

Speaker Management

  • Speakers are automatically detected and numbered (Speaker 1, Speaker 2, etc.)
  • Click the pencil icon to rename speakers (e.g., "Bob", "Alice")
  • Renamed speakers are remembered across all future recordings
  • Voice embeddings enable automatic recognition with 0.65 similarity threshold
  • "No Speaker" segments (background noise) don't have edit icons

Editing Transcripts

  • Click any segment to enter edit mode
  • Make changes in the text editor
  • Click "Save" to persist changes or "Cancel" to discard
  • Click "..." menu for additional options (Delete)

Recordings Management

  • Recordings appear in the sidebar grouped by date (Today, Yesterday, This Week, etc.)
  • Right-click any recording for options (Rename, Delete)
  • Click a recording to view its transcript
  • Active recordings show real-time elapsed timer

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments & Credits

  • FluidAudio - Excellent ASR and speaker diarization framework
  • Apple's ScreenCaptureKit - Enabling per-process audio capture
  • Apple's Accelerate framework - Optimized vector operations for embeddings
  • SwiftUI & SwiftData - Modern Apple development stack

⭐ Star this repo if you find it useful! Share it with others who need better meeting transcription.