Magic Mirror

Inspiration

Public speaking and presentation anxiety are common challenges that impact over 90% of students. We aimed to develop a tool that enables people to practice, receive feedback, and enhance their communication skills using accessible technology. The goal was to make self-improvement as simple as looking in the mirror.

What it does

Magic Mirror is a smart feedback system that uses a microphone to analyze voice tone, pace, and other metrics. Displaying real-time feedback on presentation delivery. It offers personalized improvement tips using AI-powered speech. Provides a web interface (built with Streamlit) for training exercises and performance tracking. It helps users become more confident, expressive, and effective speakers in classrooms.

How we built it

Our frontend was developed using Streamlit, enabling us to quickly prototype and design an intuitive interface that provides real-time feedback in a clean, accessible layout. This ensured users could easily interact with the system while focusing on their improvement.

For the backend, we used Python with FastAPI, which served as the core engine that connected the user interface to our AI logic and data processing. FastAPI managed all the API requests, processed audio inputs, and communicated seamlessly with AWS services.

The AWS integration played a central role in powering our backend. We used AWS AI and machine learning services — particularly for speech-to-text processing and recommendation generation. The backend would send recorded audio data to AWS for transcription and analysis, then receive structured feedback and improvement suggestions in real time.

For data management, we initially stored session logs and feedback locally in JSON files, which simplified testing and iteration. Moving forward, we plan to integrate Firebase for real-time data synchronization and user history tracking, allowing users to view progress across multiple practice sessions.

Challenges we ran into

One of the challenges we faced during the project was learning how to navigate the many tools and services within AWS. It took time to identify which features were most relevant and how to use them effectively, but this process ultimately improved our ability to integrate AWS into our workflow.

Accomplishments that we're proud of

We successfully built a functional Magic Mirror prototype capable of providing real-time feedback during presentations. This achievement demonstrated our ability to integrate multiple systems into a cohesive, interactive tool.

We designed an intuitive and accessible user interface to make presentation training simple and engaging for users. The layout emphasizes ease of use, ensuring that anyone can navigate and interpret their feedback with minimal guidance.

We developed multiple feedback categories, including voice tone, clarity, confidence, and pace. This allowed the system to deliver more personalized and meaningful performance insights to each user.

What we learned

Throughout the development of Magic Mirror, we gained valuable insight into how to integrate AI services into our product to enhance functionality and user experience. Working with tools like AWS and machine learning APIs taught us how to connect intelligent features seamlessly into our system, ensuring that feedback was both real-time and reliable.

We also recognized the importance of user-centered design when creating self-assessment tools. Our goal was to make the interface simple, intuitive, and engaging, so users could easily interpret their performance data and stay motivated to continue improving. Designing with the user in mind helped us prioritize clarity, accessibility, and emotional encouragement over technical complexity.

Finally, we learned how real-time feedback encourages consistent practice and measurable progress. By offering immediate insights on tone, clarity, and confidence, Magic Mirror transformed self-evaluation into an ongoing, data-driven process. This approach not only increased user engagement but also made skill improvement more tangible and rewarding over time.

What's next for Magic Mirror

Our next step is to launch a web and mobile app version of Magic Mirror, making real-time feedback accessible anywhere. This expansion will let users practice presentations and receive AI-driven insights across devices, with progress automatically tracked in the cloud.

We also plan to add multilingual support to make Magic Mirror accessible worldwide. Integrating natural language processing for multiple languages will help support non-English speakers and adapt to cultural nuances in communication.

Finally, we aim to implement video and emotion recognition analysis to provide deeper, more personalized feedback. By assessing facial expressions, gestures, and tone, Magic Mirror will evolve into a complete communication training tool that understands both verbal and nonverbal expressions.