Unit 2. A gentle introduction to audio applications

Welcome to the second unit of the Hugging Face audio course! Previously, we explored the fundamentals of audio data and learned how to work with audio datasets using the 🤗 Datasets and 🤗 Transformers libraries. We discussed various concepts such as sampling rate, amplitude, bit depth, waveform, and spectrograms, and saw how to preprocess data to prepare it for a pre-trained model.

At this point you may be eager to learn about the audio tasks that 🤗 Transformers can handle, and you have all the foundational knowledge necessary to dive in! Let’s take a look at some of the mind-blowing audio task examples:

• Audio classification: easily categorize audio clips into different categories. You can identify whether a recording is of a barking dog or a meowing cat, or what music genre a song belongs to.
• Automatic speech recognition: transform audio clips into text by transcribing them automatically. You can get a text representation of a recording of someone speaking, like “How are you doing today?“. Rather useful for note taking!
• Speaker diarization: Ever wondered who’s speaking in a recording? With 🤗 Transformers, you can identify which speaker is talking at any given time in an audio clip. Imagine being able to differentiate between “Alice” and “Bob” in a recording of them having a conversation.
• Text to speech: create a narrated version of a text that can be used to produce an audio book, help with accessibility, or give a voice to an NPC in a game. With 🤗 Transformers, you can easily do that!

In this unit, you’ll learn how to use pre-trained models for some of these tasks using the pipeline() function from 🤗 Transformers. Specifically, we’ll see how the pre-trained models can be used for audio classification, automatic speech recognition and audio generation. Let’s get started!