Inspiration Public speaking is a critical skill for professional and personal growth. However, many people struggle with filler words, rambling, and speaking concisely. Our team was inspired to create a solution to help people improve their public speaking abilities in a supportive environment.

What it does PartyRock is a mobile app that provides real-time feedback during speeches to help users improve their public speaking. It listens to the user's speech and provides live feedback on filler words, sentence structure, pace, tone variation, and more. PartyRock's algorithm counts filler words like "um" and "uh", highlights rambling sentences, and makes suggestions for improvement.

How we built it We built PartyRock using React Native for cross-platform iOS and Android support. The speech analysis algorithm uses machine learning models trained on hours of sample speech data. We implemented the real-time feedback system using WebSocket technology for low-latency communication.

Challenges we ran into Training accurate machine learning models for speech analysis was difficult with limited data. We iterated through multiple model architectures and data augmentation techniques. Building a real-time mobile app with WebSocket connectivity also posed synchronization issues.

Accomplishments that we're proud of We're proud to have built an app that provides actionable, supportive feedback in real-time to help people improve their speaking abilities. The speech analysis algorithm exceeded our accuracy targets once fully trained.

What we learned We learned the challenges of developing real-time mobile apps and training machine learning models with limited data. Accuracy, latency, and synchronization are all crucial for user experience.

What's next for PartyRock Next we plan to expand PartyRock's capabilities to analyze tone, emotion, facial expressions, and body language. We also aim to build group speaking practice rooms and integrate with presentation software. PartyRock has potential as a supportive public speaking aid.

Built With

  • claude
Share this project:

Updates