Moshi is a speech-text foundation model and full-duplex spoken dialogue framework. It uses Mimi, a state-of-the-art streaming neural audio codec.

Kyutai's Speech-To-Text and Text-To-Speech models based on the Delayed Streams Modeling framework.

Convolutional neural network model for video classification trained on the Kinetics dataset.

A voice chat app

Make text LLMs listen and speak

Googles NotebookLM but local

Kyutai with an "eye"

Simple high-throughput inference library

Fine-tuning Moshi/J-Moshi on your own spoken dialogue data

A text embedding extension for the Polars Dataframe library.