AI Workflow Tutorial: Custom Notebook LM

Introduction

• Purpose: Create a custom Notebook Language Model (LM) using various AI tools.
• Goal: Develop a system that produces audio summaries in a custom voice from text summaries of meetings.
• Customization: Fully customizable in terms of voice, style, and workflow.

Overview of Workflow

1. Meeting Summary to Audio File: Transform meeting summaries into audio files using a custom voice.
2. Customization: Customize prompts, voices, and styles.
3. Automation: Set up to automatically fetch files and create summaries.

Requirements

• Dropbox Account: Used for file storage and automation setup.
• Make Account: To automate the workflow, can start with a free plan.
• Blueprints: Downloadable from a public Google Drive folder for easy setup.
• 11 Labs Account: For creating a custom voice; higher quality requires a paid plan.
• Fine-tuned Model on OpenAI: Use GPT-4 fine-tuned on personal transcripts.

Step-by-Step Setup

Dropbox Setup

• Create a free account and set up a folder structure for automation:
  • Igor's Custom Meeting Summarizer
    • to do
    • in progress
    • completed summaries

Make Account Setup

• Sign up: Use the link provided for a free account.
• Import Blueprints: Import the automation blueprint for setup.

11 Labs Setup

• Custom Voice Creation:
  • Sign up for a paid plan to create a high-quality voice clone.
  • Upload a 30-minute recording of your voice.
  • Ensure no other voices are included in the recording.

Fine-Tuning Model on OpenAI

• Setup:
  • Log into OpenAI platform and ensure some credits are added.
  • Use GPT-4 for fine-tuning.
  • Prepare training data from personal transcripts.
  • Convert training data to JSONL format for upload.
  • Create the fine-tuned model.

Automation Setup

Dropbox Integration

• Link Dropbox: Connect to the to do folder to initiate automation.
• File Movement: Automate moving files from to do to in progress and completed summaries.

ChatGPT Steps

• Model Connection: Connect your fine-tuned GPT-4 model.
• Prompt Customization: Adjust context in prompts to reflect your persona.

11 Labs Integration

• API Key Setup: Connect to 11 Labs using an API key.
• Voice Selection: Choose the custom voice for output.

Final Setup

• Check and Save: Ensure all connections are correctly set.
• Run Automation: Schedule or manually run automation to test output.

Conclusion

• Customizability: The system is highly customizable, allowing for changes in voices, prompts, and file systems.
• Applications: Automated conversion of text to personalized audio summaries.
• Support: Community support is available for troubleshooting and questions.