Treble

I’ve been composing music and writing poems for several years now, but I’ve always struggled to make the accompaniments for the songs I write. The process is tedious and time-consuming, and frankly, I wouldn’t want to spend hundreds of dollars to get professional help. But what if I could get AI to do it for me? That’s how Treble was born.

​

Treble is an AI-powered musical accompaniments generating app that uses vocal inputs (like melody and lyrics) from users to create professional-grade music backing tracks in a time and cost-efficient way. It’s a great solution for artists that are non-instrumentalists, because this way they can produce music without having to pay a lot for professional orchestras.

​

I started by creating prototypes for the mobile app in Figma, following up that by researching different audio preprocessing pipelines and implementing them. The results I received looked promising, but required a lot of fine-tuning. Given my time constraints of being a university student, I pivoted and researched models that did similar things on HuggingFace. There were three that I found, and I ended up trying to implement all of them and wrote a technical product analysis evaluating the pros and cons of each ML-model in the app pipeline. I've also worked on creating a pitch deck, but all of this is still in progress as I work towards an MVP I can be proud of.