Voice2Brief

Convert voice memos into actionable briefs and meeting notes using AI transcription and processing.

This tool helps you transform audio recordings (like iOS Voice Memos) into well-structured written content using OpenAI's Whisper for transcription and large language models (GPT-4o, o1, o3-mini or Claude) for processing. Perfect for busy professionals who prefer voice recording over typing.

Features

• Audio Transcription: Convert any audio file to text using OpenAI's Whisper model
• Three Processing Modes:
  • Brief Generation: Turn voice memos about tasks into clear team briefs with assignments
  • Meeting Notes: Convert recorded meetings into structured notes with action items
  • Extended: Transform voice memos into polished, comprehensive documents
• Language Detection: Automatically detect the language of the transcription
• Structured output: Output uses sections and bullet points thanks to markdown formatting
• Flexible Output: Display results in console or save to file
• Multiple LLM Support: Compatible with OpenAI's GPT and Anthropic's Claude

Installation

The easiest way to install voice2brief is using pip

pip install voice2brief

or using pipx:

pipx install voice2brief

If you use pip, it is recommended to use a virtual environment.

Then, make sure to set your API keys in your environment. On Linux or MacOS:

export OPENAI_API_KEY=your-openai-api-key
export ANTHROPIC_API_KEY=your-anthropic-api-key

On Windows (PowerShell):

setx OPENAI_API_KEY "your-openai-api-key"
setx ANTHROPIC_API_KEY "your-anthropic-api-key"

Please note the Anthropic API key is optional and only needed to use the Claude models.

Usage

Basic usage:

voice2brief path/to/audio.m4a

Command Line Arguments:

• audio_file: Path to the audio file (required)
• --mode: Processing mode (default: brief)
  • brief: Generate a team brief with assignments
  • meeting_notes: Generate meeting notes with action items
  • extended: Create a polished document from the voice memo
• --model: LLM model to use (default: chatgpt-4o-latest)
• --output: Output file path (optional, defaults to stdout)

You can choose any model supported by OpenAI or Anthropic. Examples:

# Use the latest GPT-4o model
voice2brief audio.m4a --model chatgpt-4o-latest

# Use the o1 reasoning model
voice2brief audio.m4a --model o1-2024-12-17

# Use the o3-mini reasoning model
voice2brief audio.m4a --model o3-mini-2025-01-31

# Use the Claude Sonnet model
voice2brief audio.m4a --model claude-3-sonnet-20241022

Please note that for OpenAI models, you might need a certain tier of account to access more advanced models like o1 or o3.

Example Outputs

Brief Mode

OVERVIEW:
[A concise overview of the topic]

ASSIGNMENTS:
[List of assignments with clear ownership]

TIMELINE:
[Any deadlines or important dates]

Meeting Notes Mode

SUMMARY:
[Brief meeting summary]

KEY POINTS DISCUSSED:
[Bullet points of main discussion topics]

ACTION ITEMS:
[List of action items with owners]

DECISIONS:
[List of decisions made]

Extended Mode

[A well-structured document that preserves the speaker's intent and message,
with improved clarity and organization. The output maintains the original
tone while removing verbal artifacts and repetitions.]

Requirements

• Python 3.13+
• OpenAI API key (required)
• Anthropic API key (optional, only for Claude)

Contributing

Feel free to open issues or submit pull requests to improve the tool.

License

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