This project uses a large language model (LLM), Chat GPT 3.5 turbo to generate chord progressions and lyrics based on user input, such as mood and topic. The generated song is then sung using DiffSinger, a diffusion probabilistic model for singing voice synthesis (SVS). This software empowers educators to instantly (within a minute) create songs that enhance student learning by teaching vocabulary, concepts, rhythm, and musical skills through engaging, real-time music experiences. From the initial brainstorming to the current development, I have been responsible for the entire process. You can explore and try out the music on my YouTube channel.
Educational Song Composition using Large Language Model (LLM) with Synthesized Vocal Singing
If educators could immediatley create custom songs (complete with lyrics, melody, and vocals), it would enhance learning for young children. For example, after a lesson about addition, children could sing together "One plus one is two," reinforcing both the concept and musical development.
Nagyung Kim
(She/Her)
student, EPFL
![Tuning Hyperparameters of a Transformer [ Report ]](https://images.zapnito.com/cdn-cgi/image/metadata=copyright,fit=scale-down,format=auto,quality=95/https://images.zapnito.com/users/729637/documents/261322/eacb5ed0-ad60-4e23-84ed-ef54c8a0e91e-med.jpg)
Please sign in
If you are a registered user on Laidlaw Scholars Network, please sign in