DuoQ

Led a team of 30+ students to create an experimental voice-input, e-dating simulator for USC's capstone project class

Lead Producer and Writer

DuoQ is an e-dating sim about romancing another person through an online game. Play alongside Tala, a mysterious stranger met in the fictional FPS game To the Heart. Work (and flirt!) together using real-time voice input, solving in-game puzzles, and attempting to win over Tala’s affection.

DuoQ is a capstone project for my Advanced Games Program class that I was the lead producer of. Our game seeks to capture the magic of playing with strangers on the internet and the human compassion that drives players to play the support role in multiplayer FPS games.

Our goal was to create unconventional dating sim that feels less transactional and puts more of an emphasis on communication via our conversational AI.

My Contributions

Defined milestone goals for a year-long development process and presenting biweekly progress to faculty
Identified high-risk features with other department leads to determine feature feasibility and prioritization
Implemented a cross-discipline pod to rapidly iterate on scene perceptive game AI using GOAP
Assigned tasks using ClickUp, a project management platform like Jira
Managed V.O. recording pipeline to rapidly implement 450+ lines of V.O. in order to cover as many real world situations and possible conversation topics players may talk about
Written and edited about 100 lines of conversation, ensuring clarity, coherence, and consistent tone of character, while adhering to the project's vision and deadlines
Actively participated in brainstorming sessions and pitched different story arcs, offering valuable suggestions that lead to flirty comedy that worked with our specific voice input system
Mitigating risk of experimental features by using concentric design to align team goals and avoid high scope
Facilitated team rituals such as lead meetings, stand ups, sprint reviews, and sprint planning
Recruited, interviewed, and onboarded prospective candidates for different disciplines

How does the AI work?

Another goal of our team was to explore using emerging technologies, like large learning language models, without infringing on creative roles by generating content. We achieved this by using a combination of ChatGPT, Whisper, and Unreal's Behavior Trees.

To talk to Tala, the player uses a push to talk key to begin recording input from their microphone. When the key is released, Whisper is used to turn the audio into text. This text is then sent to ChatGPT which parses the message and sends back formatted information. This information is then fed into the Unreal AI and Tala carries out the intended action.

Page updated

Google Sites

Report abuse