🎙️ Audio Transcription Suite with Faster Whisper

--- NOTE: This proyect should be dockerized to avoid dependencies related to OS, but is not implemented yet ---

🚀 Quick Start

# List available recording devices
python src/main.py --list-devices

# Record using device 2 with large model
python src/main.py -m l -d 2

# Transcribe YouTube video with summarization
python src/main.py -l "https://youtu.be/EXAMPLE" -m l -s

🔧 Full Command Reference

Flag	Parameter	Description	Default
-m	--model	Whisper model size	m
-d	--device	Input device ID	Auto
-c	--chunk-dur	Chunk duration (seconds)	30
-l	--load	File path/YouTube URL	None
-s	--summarize	Enable GPT-4 analysis	Off

📂 File Structure

project-root/
├── audio/               # Raw recordings/chunks
├── output/              # Transcripts & summaries
├── prompts/             # Custom AI templates
└── models/              # Whisper model cache

🧠 AI Template Example

Create prompts/medical.txt:

title=Medical Analysis
Analyze patient discussion:
1. List symptoms
2. Identify medications
3. Flag concerns

[transcription here]

💻 Real-World Usage

1. Conference Recording

python src/main.py -d 4 -c 120 -s

Outputs: Timestamped transcript + bullet-point summary

2. Podcast Episode Analysis

python src/main.py -l "https://youtu.be/PODCAST_ID" -m l

Process: YouTube → WAV → 30s chunks → Full transcript

3. Historical Archive Processing

python src/main.py -l archive_tape.wav -m m -c 45

Features: Medium model, 45s chunks for context

⚠️ Requirements

1. .env file:

OPENAI_API_KEY=your_key_here

2. Hardware:

• 4GB RAM minimum
• GPU recommended for large models