Formatting Transcripts with Prompts

Welcome back! In the last lesson, you learned how to use system prompts to expand abbreviations in your audio transcriptions, making your transcripts clearer and more professional. Now, we'll build on that skill by exploring how prompts can also control the format of your transcription output. This is especially useful when you need transcripts for different business purposes, such as meeting notes or action item lists, all from the same audio file.

In this lesson, you will:

• See how prompts can shape the structure and style of your transcription output.
• Learn to write prompts that produce professional meeting notes or clear action item lists.
• Understand how to use the Whisper API to apply these formatting prompts in practice.
• Discover how to adapt your prompts for different business needs.

By the end, you'll be able to guide the AI to produce transcripts in the exact format you need, saving time and improving communication.

A system prompt is more than just a way to expand abbreviations — it's a tool for telling the AI how you want your transcript to look and feel. With the right prompt, you can:

• Set the structure (such as using bullet points or numbered lists).
• Define sections (such as "Key Discussion Points" and "Action Items").
• Control the tone (professional, casual, technical, etc.).
• Instruct the AI to extract or highlight specific information.

This is powerful because the same audio can be turned into different types of documents, depending on your needs. For example, you might want a detailed summary for your team but a simple list of tasks for your project manager. By changing the prompt, you change the output — without having to edit the transcript by hand.

Let's look at how you can use prompts to control output formatting with the Whisper API. We'll break the code into logical parts for clarity.

Now, let's see how to write prompts that produce different types of business documents.

def create_meeting_notes_prompt():
    prompt = """
    Format as professional meeting notes with expanded abbreviations:
    
    Structure:
    ## Key Discussion Points
    - Use bullet points for main topics discussed
    - Expand all abbreviations (CTO=Chief Technology Officer, API=Application Programming Interface, etc.)
    
    ## Action Items
    - Extract tasks that need to be completed
    - Include responsible parties when mentioned
    - Use bullet points starting with action verbs
    
    Use professional business language throughout.
    """
    return prompt

def create_action_items_prompt():
    prompt = """
    Extract actionable tasks from the discussion with expanded abbreviations:
    
    Instructions:
    - Convert discussions into specific action items
    - Start each item with an action verb (Complete, Develop, Implement, Review, etc.)
    - Expand all abbreviations (CTO, API, CRM, MVP, etc.)
    - Format as numbered list
    - Include responsible parties when mentioned or can be inferred
    - If no explicit tasks are mentioned, infer logical next steps from the discussion
    """
    return prompt

Here, you see two different prompts:

• The first prompt tells the AI to format the transcript as professional meeting notes, with clear sections and bullet points.
• The second prompt asks for a list of action items, using a numbered list and starting each item with an action verb.

Both prompts also remind the AI to expand abbreviations, keeping the output clear and professional.