Voice to Notes

A Python application that transcribes voice recordings to text and processes them into structured notes using AI.

Features

• Transcribe audio files to text using local Whisper model
• Process transcripts using Ollama (local LLM) to generate:
  • Key decisions and commitments
  • Current blockers and dependencies
  • Potential opportunities and ideas
  • Identified risks and concerns
  • Actionable items with priorities and due dates
• Generate consolidated daily to-do lists from multiple transcripts
• Save processed transcripts in structured JSON format

Current Status

The application currently consists of separate scripts for each step of the process. Future versions will integrate these into a more cohesive workflow.

Prerequisites

• Python 3.8 or higher
• Ollama installed and running locally
• CUDA-capable GPU recommended for faster transcription

Installation

1. Clone the repository:

git clone https://github.com/yourusername/voice-to-notes.git
cd voice-to-notes

2. Create and activate a virtual environment:

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

pip install -e .

4. Install and start Ollama:

• Follow instructions at https://ollama.ai/
• Pull the Gemma model:

ollama pull gemma3:27b

Usage

1. Place your audio files in the recordings directory.

2. Transcribe the audio files:

python scripts/transcribe_recording.py

3. Process the transcripts:

python scripts/process_transcript.py

4. Generate daily to-do lists:

python scripts/generate_daily_todos.py

The processed transcripts will be saved in the processed_transcripts directory as JSON files, and daily to-do lists will be saved in the daily_todos directory.

Project Structure

voice-to-notes/
├── recordings/           # Directory for audio files
├── transcripts/         # Directory for raw transcript text files
├── processed_transcripts/ # Directory for processed transcript JSON files
├── daily_todos/        # Directory for consolidated daily to-do lists
├── scripts/
│   ├── transcribe_recording.py  # Script for audio transcription
│   ├── process_transcript.py    # Script for transcript processing
│   └── generate_daily_todos.py  # Script for generating daily to-do lists
├── src/
│   └── voice_to_notes/
│       ├── __init__.py
│       ├── transcribe.py        # Core transcription functionality
│       └── models.py           # Pydantic models for structured output
├── pyproject.toml
└── README.md

Output Format

Processed Transcripts

Processed transcripts are saved as JSON files with the following structure:

{
  "summary": {
    "key_decisions": ["Decision 1", "Decision 2"],
    "blockers": ["Blocker 1", "Blocker 2"],
    "opportunities": ["Opportunity 1", "Opportunity 2"],
    "risks": ["Risk 1", "Risk 2"]
  },
  "action_items": [
    {
      "description": "Action item description",
      "priority": 3,
      "due_date": "2024-03-20T10:00:00",
      "blockers": ["Dependency 1"],
      "status": "pending"
    }
  ],
  "metadata": {
    "processed_at": "2024-03-19T15:30:00",
    "model_used": "ollama",
    "version": "1.0"
  }
}

Daily To-Do Lists

Daily to-do lists are saved as JSON files with the following structure:

{
  "date": "2024-03-20",
  "action_items": [
    {
      "description": "Action item description",
      "priority": 3,
      "due_date": "2024-03-20T10:00:00",
      "blockers": ["Dependency 1"],
      "status": "pending"
    }
  ],
  "key_decisions": ["Decision 1", "Decision 2"],
  "blockers": ["Blocker 1", "Blocker 2"],
  "opportunities": ["Opportunity 1", "Opportunity 2"],
  "risks": ["Risk 1", "Risk 2"]
}

License

MIT License

Scripts

1. transcribe_recording.py

Transcribes audio recordings into text files.

2. process_transcript.py

Processes transcript text files to extract actionable insights using Ollama.

3. generate_daily_todos.py

Aggregates processed transcripts by date and generates consolidated daily to-do lists using Ollama.