Songfinch, recognized for artist-written and produced custom songs, sought to further personalize their offerings with Starring You, a feature that allows users to become the voice of any of their previously purchased Instant songs.

Leveraging Large Language Models (LLMs), Starring You analyzed user-recorded voice data across various emotional expressions and learned to generate new speech – replacing the original artist's voice with the user’s.

I designed the beta release which guided users through recording themselves speaking a set of sentences designed to evoke 5 distinct emotions – joy, sadness, anger, fear, and disgust. This process captured the nuanced vocal data required for the AI to seamlessly integrate their voice into the song.