Two undergrads built an AI speech model to rival NotebookLM

Two undergraduate students have developed Dia, an open-source AI model that generates podcast-style audio clips. The model, inspired by Google's NotebookLM, allows users to customize voices, insert nonverbal cues, and even clone voices. Trained using Google's TPU Research Cloud, Dia can run on most modern PCs. While the technology shows promise, it lacks safeguards against potential misuse for disinformation or scams. The creators plan to build a synthetic voice platform with social features based on Dia and future models.