NotebookLM with Gemini and Elevenlabs (Detailed Documentation)

If you’re tech-savvy or active on Twitter, you’ve probably heard about Google NotebookLM — Google’s AI-powered research assistant. It can analyze uploaded documents (Google Docs, PDFs, and URLs) to generate summaries, answer questions, and create insights. One unique feature of NotebookLM is the ability to transform documents into deep-dive conversational podcasts.

In this guide, we’ll build NotebookLM (PDF to podcast) using Gemini for dialogue generation and Elevenlabs for text-to-speech. Let’s dive right into it!

App Demo

1. Dialogue Generation Using Gemini

>Create Your Gemini API Key

To start, we’ll need to create a Gemini API key from Google AI Studio.

1. Navigate to Google AI Studio and sign in or create an account.
2. Once logged in, click on Get API Key to generate your API key.

Create Gemini API Key

Now, store this API key in your .env file as:

GEMINI_API_KEY="YOUR_GEMINI_API_KEY"

With this setup, we’re ready to integrate Gemini for dialogue generation.

In the gemini.py file of our project, we use the API key to call the Gemini API and create a conversation based on your PDF input. This is the first step in our process of converting PDFs into podcasts.

2. Text-to-Speech Using Elevenlabs

Why Elevenlabs?

Elevenlabs offers over 3,000 voices across 32 languages, providing great flexibility in voice generation. Let’s use Elevenlabs to convert the generated dialogue from Gemini into realistic podcast audio.

Create Your Elevenlabs API Key

1. Go to Elevenlabs and create your account.
2. Once logged in, navigate to the Text-to-Speech App section here.
3. Click on your profile icon, then go to the API Key section, and create your API key.

Create Elevenlabs API Key

Set the ELEVENLABS_API_KEYin your .env file:

ELEVENLABS_API_KEY="YOUR_ELEVENLABS_API_KEY"

To ensure everything is working, you can test Elevenlabs’ Text-to-Speech API using the following curl command:

curl -X POST https://api.elevenlabs.io/v1/text-to-speech/<voice_id> \
     -H "xi-api-key: <apiKey>" \
     -H "Content-Type: application/json" \
     -d '{
  "text": "text"
}'

Replace voice_id and apiKey with your credentials.

Enabling Voices

To select a voice for your podcast:

1. Go to Elevenlabs Voices.
2. Select the voice you want and click on the view button.
3. Copy the voice ID.

Copy Voice ID

Once you have the voice ID, you can specify which voice (male or female) you want for the dialogue speakers in your elevenlabs.py file.

Putting It All Together

Now that we’ve set up both Gemini for generating dialogue and Elevenlabs for converting the dialogue into speech, it’s time to integrate them.

The process works in two steps:

1. Dialogue Generation: We use Gemini to analyze the PDF and create a conversational dialogue between two speakers (mentor and student).
2. Text-to-Speech: We then use Elevenlabs to convert that dialogue into audio files, which you can listen to as a podcast.

Let’s start Project Installation:

Project Installation

Follow these steps to get the project up and running:

1. Clone the Repository

git clone https://github.com/chiragjoshi12/pdf-to-podcast.git
cd pdf-to-podcast

2. Install Dependencies

pip install -r requirements.txt

3. Set Up API Keys

Create a .env file in the root directory and add your Gemini and Elevenlabs API keys:

GEMINI_API_KEY="YOUR_GEMINI_API_KEY"
ELEVENLABS_API_KEY="YOUR_ELEVENLABS_API_KEY"

4. Run the Application

Finally, start the application using Streamlit:

streamlit run main.py

The app should now be live on your local machine, and you can start uploading PDFs to convert them into podcasts.

This is my first attempt at writing a technical blog, so your feedback and suggestions are greatly appreciated as I continue to learn and improve!

Thanks 🙏 !