Inspiration

I joined this hackathon with the idea of creating a voice-powered AI Chat App, Nova, that could run on Android devices and work in the background—similar to virtual assistants like Alexa and Siri. My goal was to make a hands-free conversational AI that people could interact with anytime. Even with limited time and resources, the challenge inspired me to see how far I could go.

What it does

Nova is a voice-activated AI chat app for Android that allows users to have human-like conversations by speaking naturally. Running seamlessly in the background, it listens, processes, and responds in real-time, so users can interact with it hands-free. Nova transcribes user audio, converts it to text for processing, and then delivers the response as audible speech, providing a conversational experience that feels truly interactive.

How we built it

Nova was built from scratch in Java using Android SDK for core functionality and Retrofit for network calls. I used GPT to handle the language processing, transcribing voice inputs to text and generating natural language responses. Due to time constraints, I initially focused on integrating pre-built components and libraries to get a working prototype. However, I had to balance speed and accuracy, so I later incorporated additional tools for smoother audio handling and user experience.

Challenges we ran into

One of the biggest challenges was building Nova on an Android device without a PC and also I am the only person that does the job in that 4days. This limited my ability to debug quickly and use certain tools, and I faced memory management issues that led to some bugs, such as memory leaks, anyway there are still some memory leaks in the code. Additionally, creating a voice-powered AI that continuously runs in the background required optimizing the app to minimize battery and memory usage, which proved challenging within the hackathon’s timeframe.

Accomplishments that we're proud of

I'm proud to have built a functional prototype within just four days, especially given the hardware limitations. Getting Nova to successfully listen, process, and respond to user queries in real-time is a milestone. Completing this project solo and overcoming technical and hardware constraints has been rewarding.

What we learned

This project taught me a lot about the challenges of real-time audio processing and efficient memory management on Android. I learned how to optimize network calls, manage limited resources, and navigate the challenges of voice-powered AI development on mobile. I also gained valuable experience in problem-solving under pressure, which has strengthened my skills in mobile app development.

What's next for Nova

Next, I plan to improve Nova’s stability by addressing the memory leaks and refining the app's performance. I’d also like to add more personalization options, such as customizable voice responses and user-defined triggers. Additionally, enhancing compatibility with a wider range of Android devices and potentially developing a web version are also on the roadmap.

Built With

  • accessibilityservice
  • audiorecorder
  • gpt
  • java
  • speechrecognizerandroid
  • xml
Share this project:

Updates