Interactive-Podcasts

Enhanced Listening: Interactive Podcasts. This tool uses AI to transform standard audio files into active learning experiences. It automatically generates and inserts comprehension questions, utilizing active recall to improve retention.

Files

• transcriber.py: transcribes the input podcast using OpenAI's Whisper API. Outputs transcript.json containing the full text and a word-level list with start and end timestamps.

• question_generator.py: reads transcript.json and prompts GPT-4o-mini to generate comprehension questions placed coherent points in the podcast. Outputs questions.json with each question, its answer, and the word index after which it should be inserted.

• text_to_speech.py: reads questions.json and generates an MP3 audio file for each question using OpenAI's TTS API (tts-1, voice: alloy). Updates questions.json with the path to each generated audio file and the timestamp after which it should be inserted based on the word index provided by question_generator.py.

• audio_editor.py: assembles the final podcast by splitting the original audio at each insertion timestamp and adding the question audios. Adds a bell sound before each question, a 6-second silence for the listener to think, and another bell when the pause ends. Outputs the final file to output/episode_interactive.mp3.

• run_pipeline.py: runs all four steps in sequence. See usage instructions below.

• analysis/analysis.py , analysis/visualizations.ipynb and analysis/CS_1377_Final_Visualizations.ipynb: contains the statistical analysis for the experiment testing the tool's effectiveness on information retention, including ANOVA, t-tests, and performance breakdowns by group and question format and their corresponding visualizations.

Install Dependencies

run: pip install -r requirements.txt

Configure environment

Create a .env file in the root of the project and add you OpenAI API key:

OPENAI_API_KEY=your_openai_api_key_here

Usage

Place your podcast file in the project root and rename it episode.mp3 (or update the filename in the scripts).

Run the full pipeline: run_pipeline.py

This will run all four steps in order and output the final file to output/episode_interactive.mp3.

Run steps individually:

• python transcriber.py
• python question_generator.py
• python text_to_speech.py
• python audio_editor.py

Sounds:

Twinkling Stars Sound Effect https://www.youtube.com/watch?v=okMJSAxfOzc

Ding - Sound Effect https://www.youtube.com/watch?v=Fu82s5DnhBc