A Korean-based startup called Nari Labs has released an open-source AI model that can generate podcast-style clips similar to Google’s NotebookLM. The model, named Dia, uses 1.6 billion parameters and can be run on most modern PCs.
Dia allows users to customize speakers’ tones and insert disfluencies, coughs, laughs, and other nonverbal cues. It can also clone a person’s voice. According to TechCrunch, the quality of the voices is competitive with other tools available.
The model was trained using Google’s TPU Research Cloud program, which provides researchers with free access to the company’s AI chips. However, there are concerns about the lack of safeguards and data scraping practices used during training.
Nari Labs plans to create a synthetic voice platform with a “social aspect” on top of Dia and larger models in the future. The company also intends to release a technical report for Dia and expand its support to languages beyond English.
Source: https://techcrunch.com/2025/04/22/two-undergrads-built-an-ai-speech-model-to-rival-notebooklm