Repolode
Submit
facebookresearch

facebookresearch/AugLy

A Python data augmentation library for audio, image, text and video, with over 100 transforms, including social-media-style edits for robustness testing.

View on GitHubWebsite
Stars
5,100
Forks
310
Open issues
27
Last commit
18 days ago

What is it?

AugLy is a Python library for data augmentation across four modalities: audio, image, text and video. Each modality has its own sub-library with more than 100 augmentations in total. The transforms come as both function-based and class-based versions, along with composition operators, and can optionally return metadata describing the applied transform and its intensity.

Many of the augmentations are modeled on what users do on internet platforms, such as turning an image into a meme, overlaying text or emojis on images and videos, or reposting a screenshot from a social feed. The library also includes more generic transforms. It is aimed at training models or measuring their robustness on tasks like copy detection, hate speech detection and copyright infringement, where this kind of content is common.

AugLy requires Python 3.6 or later. It installs with pip, and each sub-library can be installed separately through extras such as augly[audio]. The project also ships emoji, font and screenshot template assets for use with certain augmentations.

Key features

  • 01Covers audio, image, text and video augmentations in separate sub-libraries.
  • 02Offers over 100 augmentations, including meme-style and screenshot-style edits.
  • 03Provides both function-based and class-based transforms with composition operators.
  • 04Can return metadata describing each applied transform and its intensity.
  • 05Supports per-modality installation via pip extras.

Use cases

  • Augmenting training data for copy detection or near-duplicate detection models.
  • Testing how robust a model is to social-media-style image and video manipulations.
  • Generating augmented text or audio samples for content moderation research.

Quick start

  1. 1

    Install the full AugLy package from PyPI

    pip install augly[all]
  2. 2

    Or install only the dependencies for one sub-library, such as audio

    pip install augly[audio]

Things to know

  • The README says installs may need extra system packages for python-magic, depending on the environment.
  • Some dependencies may be licensed under different terms than AugLy's MIT license.
  • The license metadata is listed as NOASSERTION, so the license should be checked directly in the repository.

In collections

Computer vision libraries & toolsNatural language processing tools & libraries

Topics

#data-augmentation#machine-learning#computer-vision#nlp#audio#video#robustness#python

FAQ

What is AugLy?
A Python data augmentation library for audio, image, text and video, with over 100 transforms, including social-media-style edits for robustness testing. AugLy is a Python library for data augmentation across four modalities: audio, image, text and video. Each modality has its own sub-library with more than 100 augmentations in total. The transforms come as both function-based and class-based versions, along with composition operators, and can optionally return metadata describing the applied transform and its intensity.
Is AugLy free to use? What license does it have?
AugLy is open source under the Other license. Check the license terms before using it in commercial products.
Is AugLy still maintained?
Actively maintained. The last commit to AugLy was on Oct 20, 2026.
What can I use AugLy for?
Augmenting training data for copy detection or near-duplicate detection models. Testing how robust a model is to social-media-style image and video manipulations. Generating augmented text or audio samples for content moderation research.
How do I install AugLy?
Install the full AugLy package from PyPI: pip install augly[all] Or install only the dependencies for one sub-library, such as audio: pip install augly[audio]
Where can I get AugLy?
The source code and installation instructions are on GitHub: https://github.com/facebookresearch/AugLy

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects