Smart Voice Summarizer

• 

  Your one stop shop for summaries

Inspiration

As a student, it is very time consuming to sit and watch all recorded lectures. We wanted to aim at something that is simple, time efficient and accurate. Then we realized that this is an issue that people from all sectors face. This is what motivated us to come up with our problem statement.

What it does

It takes raw audio input, removes noise and transcribes it. It can then summarize it and convert the summary text to speech using Neuphonic's API.

How we built it

• Python
• Google Colab / VS Code
• Whisper for transcription
• Neuphonic for voice synthesis
• noisereduce / DeepFilterNet for noise cleaning
• Streamlit for frontend

Challenges we ran into

• Cleaning audio files
• Lack of free APIs
• Inaccurate transcription

Accomplishments that we're proud of

• Successfully integrating Neuphonic's API.
• Finding alternatives to training models in a restricted timeframe

What we learned

We learned how to integrate different APIs in web applications and use tech stacks that we were not familiar with like Streamlit.

What's next for Smart Voice Summarizer

Due to constraints on time and resources, these are what we would like to implement later on:

• More efficient noise removal/reduction
• Quicker and more precise transcription
• Customizable summary length ( audio and text)

Built With

• apis
• machine-learning
• python
• streamlit