Archer VCS

About this project
Working on this project alongside Archer Pechawis, a Cree artist and educator, has made the idea of musical collaboration feel especially meaningful. The goal is to build a system that can listen to Archer singing and respond musically in a way that feels connected to the performance, rather than simply play a predetermined output.
The core audio pipeline captures a singer's voice, estimates pitch using methods such as YIN and RMVPE, and analyses rhythm and phrasing to help decide how the system should respond. A harmony engine selects musical responses, from unison and intervals to drones and call-and-response, and a digital signal processing pipeline renders them into sound. I've also worked with an optional RVC voice-conversion component, which runs separately so its heavier processing does not have to block the main audio pipeline.
The project extends beyond music, too. Its research application brings together language and audio tools, including Cree morphological analysis, Cree-to-English translation, transcription, live captions, sentiment analysis, pitch analysis, and a notes-grounded chat assistant. Connecting these components has taught me to think about a system as a whole, not just as a collection of models that work individually.
I've been fortunate to explore these questions through collaborative research with people whose perspectives and knowledge differ from my own. That experience has helped me appreciate how much there is to learn when technical work meets a real creative and cultural context. Archer is still an evolving project, and that is part of what I enjoy about it: there are always new questions about responsiveness, interaction, and what a useful musical partner could become.