torchaudio

Warning

Starting with version 2.8, we are refactoring TorchAudio to transition it into a maintenance phase. As a result:

• Most APIs listed below are deprecated in 2.8 and will be removed in 2.9.

• The decoding and encoding capabilities of PyTorch for both audio and video are being consolidated into TorchCodec. For convenience, we provide load_with_torchcodec() as a replacement for load() and save_with_torchcodec() as a replacement for save(), but we recommend that you port your code to native torchcodec APIs.

Please see https://github.com/pytorch/audio/issues/3902 for more information.

I/O

torchaudio top-level module provides the following functions that make it easy to handle audio data.

info	DEPRECATED
load	Load audio data from source.
load_with_torchcodec	Load audio data from source using TorchCodec's AudioDecoder.
save	Save audio data to file.
save_with_torchcodec	Save audio data to file using TorchCodec's AudioEncoder.
list_audio_backends	DEPRECATED

Backend and Dispatcher

Decoding and encoding media is highly elaborated process. Therefore, TorchAudio relies on third party libraries to perform these operations. These third party libraries are called backend, and currently TorchAudio integrates the following libraries.

Please refer to Installation for how to enable backends.

Conventionally, TorchAudio has had its I/O backend set globally at runtime based on availability. However, this approach does not allow applications to use different backends, and it is not well-suited for large codebases.

For these reasons, in v2.0, we introduced a dispatcher, a new mechanism to allow users to choose a backend for each function call.

When dispatcher mode is enabled, all the I/O functions accept extra keyward argument backend, which specifies the desired backend. If the specified backend is not available, the function call will fail.

If a backend is not explicitly chosen, the functions will select a backend to use given order of precedence and library availability.

The following table summarizes the backends.

Priority	Backend	Supported OS	Note
1	FFmpeg	Linux, macOS, Windows	Use get_audio_decoders() and get_audio_encoders() to retrieve the supported codecs. This backend Supports various protocols, such as HTTPS and MP4, and file-like objects.
2	SoX	Linux, macOS	Use list_read_formats() and list_write_formats() to retrieve the supported codecs. This backend does not support file-like objects.
3	SoundFile	Linux, macOS, Windows	Please refer to the official document for the supported codecs. This backend supports file-like objects.

Dispatcher Migration

We are migrating the I/O functions to use the dispatcher mechanism, and this incurs multiple changes, some of which involve backward-compatibility-breaking changes, and require users to change their function call.

The (planned) changes are as follows. For up-to-date information, please refer to https://github.com/pytorch/audio/issues/2950

• In 2.0, audio I/O backend dispatcher was introduced. Users can opt-in to using dispatcher by setting the environment variable TORCHAUDIO_USE_BACKEND_DISPATCHER=1.

• In 2.1, the disptcher became the default mechanism for I/O.

• In 2.2, the legacy global backend mechanism is removed. Utility functions get_audio_backend() and set_audio_backend() became no-op.

Furthermore, we removed file-like object support from libsox backend, as this is better supported by FFmpeg backend and makes the build process simpler. Therefore, beginning with 2.1, FFmpeg and Soundfile are the sole backends that support file-like objects.