Meeting Diary

A powerful CLI tool to transcribe and diarize audio/video files using AssemblyAI. Automatically identifies speakers and generates transcripts in multiple formats.

Features

• 🎙️ Automatic speaker diarization
• 👥 Interactive speaker identification with context
• 📝 Multiple output formats (Markdown, SRT, TXT, JSON)
• 🕒 Timestamps for each segment
• 🔑 Secure API key management
• 💾 Smart caching for faster processing
• 💻 Cross-platform support

Installation & Usage

Quick Start (Recommended)

You can use meeting-diary directly without installation using npx or bunx:

# Using npx (Node.js)
npx meeting-diary input.mp4

# Using bunx (Bun)
bunx meeting-diary input.mp4

Global Installation (Alternative)

If you prefer to install the tool globally:

# Using npm
npm install -g meeting-diary

# Using yarn
yarn global add meeting-diary

# Using bun
bun install -g meeting-diary

Then use it as:

meeting-diary input.mp4

Usage

Basic Usage

meeting-diary input.mp4

This will:

1. Transcribe and diarize your audio/video file
2. Help you identify each speaker by showing their most significant contributions
3. Generate a timestamped transcript in markdown format

Output Formats

meeting-diary input.mp4 -f txt  # Simple text format
meeting-diary input.mp4 -f srt  # SubRip subtitle format
meeting-diary input.mp4 -f json # JSON format with detailed metadata
meeting-diary input.mp4 -f md   # Markdown format (default)

Markdown Format (Default)

The markdown format includes:

• Timestamp for each segment
• Speaker list
• Chronological transcript with speaker attribution
• Processing metadata

Example:

# Meeting Transcript

_Processed on 2/10/2024, 3:43:26 PM_
_Duration: 5 minutes_

## Speakers

- **Hrishi**
- **Alok**

## Transcript

[0:00] **Hrishi**: Yeah, didn't have a chance yet...
[0:15] **Alok**: No engagement in terms of my Mushroom photos.
[0:18] **Hrishi**: Basically Samsung phones have the ability...

Speaker Identification

You can identify speakers in two ways:

1. Interactive identification (default):

meeting-diary input.mp4

The tool will:

• Show you the most significant contributions from each speaker
• Display context (what was said before and after)
• Show previously identified speakers for context
• Ask you to identify each speaker in turn

2. Specify speakers up front:

meeting-diary input.mp4 -s "John Smith" "Jane Doe"

All Options

Options:
  -o, --output <file>     Output file (defaults to input file name with new extension)
  -f, --format <format>   Output format (json, txt, srt, md) (default: "md")
  -s, --speakers <names>  Known speaker names (skip interactive identification)
  --skip-diarization     Skip speaker diarization
  -v, --verbose          Show verbose output
  --api-key <key>        AssemblyAI API key (will prompt if not provided)
  --no-cache            Disable caching of uploads and transcripts
  --cache-dir <dir>     Directory to store cache files
  --no-interactive      Skip interactive speaker identification
  -h, --help             display help for command

Caching

The tool automatically caches uploaded audio files and transcripts to avoid unnecessary re-processing. This is especially useful when:

• Experimenting with different output formats
• Re-running transcription with different speaker names
• Processing the same file multiple times

Cache files are stored in your system's temporary directory by default. You can:

• Disable caching with --no-cache
• Change cache location with --cache-dir
• Cache is enabled by default for faster processing
• Cache files are automatically cleaned up by your OS's temp file management

API Key

You'll need an AssemblyAI API key to use this tool. You can:

1. Set it as an environment variable: ASSEMBLYAI_API_KEY=your-key
2. Pass it via the command line: --api-key your-key
3. Let the tool prompt you for it (it can be saved for future use)

Development

# Clone the repository
git clone https://github.com/southbridgeai/meeting-diary.git
cd meeting-diary

# Install dependencies
bun install

# Build
bun run build

# Run tests
bun test

# Development mode
bun run dev

License

Apache-2.0 - see LICENSE for details.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.