# pytorch/audio

Data manipulation and transformation for audio signal processing, powered by PyTorch

Repository: https://github.com/pytorch/audio
Canonical: https://ross.abutalabs.com/products/audio
Homepage: https://pytorch.org/audio
Language: Python
License: BSD-2-Clause
License Family: permissive
Topics: audio, python, io, speech, machine-learning, pytorch, audio-processing
Last push: 2026-08-26T11:42:17+00:00

## Health v2 (maintenance only)
Score: 87/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 64, longevity 100
- inputs: {"age_days": 3408, "days_push": 7, "days_rel": 163, "gap_med": 62, "n_releases_24m": 10}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2927, forks 793 (observed 2026-08-28T04:07:30.483401+00:00)

## What it is
TorchAudio is PyTorch's audio library providing data manipulation, transforms, and dataset loaders for audio and speech machine learning. It is now in maintenance mode, tightly scoped to GPU-accelerated audio processing for ML pipelines.

## Use cases
- compute mel spectrograms and MFCCs for audio models
- load and preprocess speech datasets for training
- resample and encode audio as tensors in PyTorch
- run forced alignment for speech recognition
- match Kaldi fbank features in PyTorch

## When to choose
- you train audio or speech models with PyTorch and need GPU-accelerated audio transforms
- you want dataset loaders and feature extraction consistent with torch tensors and autograd

## When to avoid
- you need a general-purpose signal processing library outside machine learning
- you rely on features deprecated in 2.8 and removed in 2.9, such as I/O and decoding utilities

## Facets
- artifact type: library
- maturity: maintenance
- function: audio-processing, machine-learning, speech-recognition
- domain: machine-learning, speech-processing, deep-learning
- platform: python, cross-platform
- tags: pytorch, torchaudio, audio-transforms, spectrogram, dataloaders, signal-processing, audio, gpu

## Member repositories
- pytorch/audio (main) score 87

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:30.483401+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:33:35.980852+00:00, confidence not recorded.
  - readme: https://github.com/pytorch/audio (fetched 2026-08-28T04:07:30.483401+00:00, sha 35a5b153e0dc)
  - homepage: https://pytorch.org/audio (fetched 2026-08-29T09:48:48.724057+00:00, sha 21ac2b9ae40a)
- Data as of 2026-08-30T08:39:29.467469+00:00.