-
Intro Lobby Screen
-
Role Selection Page: Detective
-
Gameplay Screen
-
Role Selection Page: Mafia
-
Background!
Demo
Inspiration
We were inspired by those late nights with friends where all you want is to relax and enjoy some games. However, there are always times when you want to play a game like Mafia and don't have enough people, or the bots in game feel repetitive with scripted voice lines. We wanted to tackle this problem by creating AI players that aren't just fillers but actively engage in conversations and react naturally to in-game events. Our goal is to use AI and voice technology to foster deeper connections between players, whether human or AI, and bring back the joy of shared gaming experiences
What it does
Our project, MAIfia, brings the classic social deduction game Mafia to life with an AI twist. It allows human and AI players to interact in real-time using voice chat. AI players take on the roles of Mafia, Detective, Doctor, Mayor, and Civilians, creating an immersive experience where players engage in turn-based conversations powered by AWS technologies like Polly and Bedrock. This creates a seamless, fun environment for bluffing, deduction, and strategic gameplay.
How we built it
Backend: Our backend was developed using Python with Flask, leveraging React WebSocket for real-time interactions. We used Amazon Polly for AI-generated voices, Amazon Bedrock to access generative AI models, Amazon Transcribe to convert speech to text, and Amazon S3 for data storage.
Frontend: Our frontend was built with a variety of React components and CSS files, which interact with the Python backend to handle game logic and visually display in-game events.
Challenges we ran into
There were many issues with multiplayer connection for our game, that was the hardest part to implement. We also struggled with who was allowed to speak and when.
Accomplishments that we're proud of
We’re proud of successfully creating an AI-driven Mafia game with voice interaction, something that adds a unique spin to the traditional game. We managed to create an AI player that can listen to in-game discussion and react to it accordingly.
What we learned
This project was an incredible opportunity to explore a wide range of technologies that were new to us. We gained hands-on experience with tools like AWS Bedrock, Amazon Polly, Amazon Transcribe, and React WebSocket, developing a deep understanding of their strengths and limitations. This challenge has significantly expanded our technical toolkit, equipping us for future projects. More importantly, we discovered that with focus and collaboration, we can build something truly impressive in just 24 hours. The experience was both inspiring and a testament to what can be achieved under tight deadlines.
Log in or sign up for Devpost to join the conversation.