Automated Meeting Minutes from Audio

Transcription is only half the battle. We'll use OpenAI's Whisper to convert speech to text, and then GPT-4 to summarize it into actionable bullet points.

The Pipeline 1. Whisper: Transcribe `meeting.mp3`. 2. GPT-4: Distill 60 minutes of talk into 5 key points.

import openai

# 1. Transcribe
audio_file = open("meeting.mp3", "rb")
transcript = openai.Audio.transcribe("whisper-1", audio_file)

# 2. Summarize
response = openai.ChatCompletion.create(
    model="gpt-4",
    messages=[
        {"role": "system", "content": "You are a professional secretary."},
        {"role": "user", "content": f"Summarize this: {transcript['text']}"}
    ]
)
print(response.choices[0].message.content)