Inspiration Public speaking anxiety affects millions of people worldwide, yet personalized coaching remains expensive and inaccessible. We were inspired to democratize public speaking training by leveraging AI technology to provide instant, objective feedback that helps people improve their communication skills anytime, anywhere.
What it does SpeakFlow is an AI-powered public speaking coach that:
Provides real-time feedback on body language, voice tone, and speaking pace Records and analyzes practice sessions with instant, actionable insights Organizes recordings by topics for structured practice Generates detailed performance reports Offers secure cloud storage for reviewing past sessions Creates a personalized learning journey through data-driven feedback How we built it We created a full-stack application using:
Frontend: Next.js with TypeScript, Tailwind CSS, and Shadcn UI components Backend: FastAPI (Python) for real-time WebSocket communication AI Analysis: Custom feedback agent using Google's Generative AI Audio Processing: Librosa for voice analysis Video Processing: Real-time frame capture and analysis Database: Supabase for authentication and data storage Cloud Storage: Supabase Storage for video files Real-time Communication: WebSocket for instant feedback Challenges we ran into Implementing smooth real-time video processing and feedback without latency Balancing AI analysis accuracy with performance Managing WebSocket connections and state across multiple sessions Handling large video files efficiently Coordinating frontend and backend communication for seamless user experience Optimizing AI feedback to be both meaningful and actionable Accomplishments that we're proud of Built a robust real-time feedback system that processes video and audio simultaneously Created an intuitive user interface for recording and reviewing sessions Implemented secure user authentication and data storage Developed an AI system that provides contextual and actionable feedback Successfully handled real-time data streaming and processing Created a scalable architecture that can handle multiple concurrent users What we learned Advanced WebSocket implementation for real-time communication Video and audio processing techniques in Python Integration of AI models for real-time analysis Full-stack development with Next.js and FastAPI Cloud storage and database management Real-time data streaming and processing User authentication and security best practices
Log in or sign up for Devpost to join the conversation.