What is notebook?

This repo is a collection of interactive tutorials and demos for ESPnet, an open-source toolkit for speech processing. It teaches you how to build and use systems that can transcribe spoken audio into text, generate synthetic speech from written words, translate spoken language, and clean up noisy recordings. The notebooks are organized into a few categories. Demo notebooks let you run pre-trained models right in your browser to see things like real-time speech recognition or text-to-speech generation. Training notebooks show how to fine-tune existing models on your own data using a simplified interface called ESPnet-EZ. There are also full course materials from Carnegie Mellon University classes that walk through speech processing concepts step by step, covering topics like speaker recognition, speech translation, and speech enhancement. The audience is primarily researchers, students, and developers who want to experiment with speech technology without starting from scratch. A startup founder building a transcription app could use the demos to test how well different models perform on their audio. A graduate student could follow the course notebooks to learn how speech recognition pipelines work end to end. Someone with a specific use case, like transcribing a regional dialect, could use the fine-tuning tutorials to adapt a pre-trained model to their own dataset. The project leans heavily on Jupyter notebooks, which mix explanatory text with runnable code, making it approachable for people who learn best by doing. It also includes utilities like converting models into ONNX format, a standard that makes it easier to deploy models in production environments outside of the research toolkit.