It's Not Artificial (C'est Pas Artificiel)

Homage to "C'est pas Sorcier"

This project is an homage to the classic French educational TV show, "C'est pas Sorcier". The show was famous for making complex scientific topics simple, engaging, and fun for young audiences.

This script aims to capture that spirit of curious learning by using generative AI.

🚀 Description

This Google Colab notebook uses the Google Gemini API to generate a short, educational audio dialogue between two speakers. You can customize the theme, the target age for the content, the language, and even the voices of the speakers.

The script first generates a text transcript where one speaker asks questions and the other, an expert, provides clear answers.

It then uses a multi-speaker text-to-speech model to create a high-quality audio file of the conversation, bringing the educational script to life.

✨ Features

• Dynamic Content Generation: Creates educational dialogues on any theme you can imagine, from black holes to MLOps.

• Multi-Speaker Audio: Leverages Gemini's advanced multi-speaker Text-to-Speech capabilities for a natural-sounding and engaging conversation.

• Highly Customizable: Easily change the parameters in the "Settings" section of the notebook to tailor the output.

⚙️ How to Use

1. Open in Google Colab: Click the "Open In Colab" badge at the top of this README.

2. Add Your API Key: You will need a Google Gemini API Key. You can get one for free from Google AI Studio.

   • In the Colab notebook, find the API_KEY field and paste your key into the quotes.

3. Adjust the Settings: Modify the parameters like THEME, AGE, and LANG to fit your needs.

4. Run the Notebook: Execute the cells in order from top to bottom. The final cell will generate and play the audio file.

🔧 Settings

You can customize the following parameters at the top of the notebook:

• API_KEY: Your secret Google Gemini API key.

• AGE: The target age for the content, which influences the complexity of the dialogue.

• LANG: The language of the dialogue (currently supports "English" or "French").

• THEME: The scientific or technical topic for the conversation.

• MINUTES: The approximate length of the generated audio dialogue.

• SPEAKER_1_NAME / SPEAKER_2_NAME: The names of the two speakers in the transcript.

• SPEAKER_1_VOICE / SPEAKER_2_VOICE: The pre-built voice models to use for each speaker.

• TRANSCRIPT_MODEL: The Gemini model used for generating the text.

• TEXT_TO_SPEECH_MODEL: The Gemini model used for generating the audio.