Release v0.43.0

Author: iver56

README.md

@@ -91,21 +91,23 @@ The API documentation, along with guides, example code, illustrations and exampl
 
 # Changelog
 
-## [0.42.0] - 2025-07-04
+## [0.43.0] - 2025-09-09
 
-### Added
+### Changed
 
-* Add support for Python 3.13
-* Add support for librosa 0.11.0
+* Make `LoudnessNormalization` **400% faster** by switching from pyloudnorm to the new and faster alternative [loudness](https://github.com/iver56/loudness)
+* Improve `AddBackgroundNoise` performance: Only load/decode the part of the noise file that is needed, instead of the entire file.
+* Bump min Python version to 3.10
+* Switch to a more modern type hint style
 
-### Changed
+### Deprecated
 
-* Make `Mp3Compression` **25-300% faster** (depending on hardware, audio properties like duration and number of channels and various params, like bitrate) with the new `backend="fast-mp3-augment"` (now default). The extra dependency for this is [fast-mp3-augment](https://github.com/iver56/fast-mp3-augment), which pulls a few useful tricks for faster execution.
-* Make `Limiter` **30% faster** and easier to install (extra dependency is now [numpy-audio-limiter](https://github.com/iver56/numpy-audio-limiter) instead of [cylimiter](https://github.com/pzelasko/cylimiter/)). The `Limiter` behavior has not changed, although there are minor numerical differences.
+* Deprecate `"pydub"` backend in `Mp3Compression`, because pydub isn't maintained anymore, and depends on audioop, which got removed in Python 3.13. The recommended alternative is `"fast-mp3-augment"`
 
-### Fixed
+### Removed
 
-* Handle non-contiguous audio ndarray input to `PitchShift` and `TimeStretch` properly
+* Remove upper Python version limit to avoid hindering early adopters
+* Remove LRU cache in `AddBackgroundNoise`, due to the changed way of loading audio, and because such audio datasets are commonly larger than RAM anyway
 
 For the full changelog, including older versions, see [https://iver56.github.io/audiomentations/changelog/](https://iver56.github.io/audiomentations/changelog/)
 

audiomentations/__init__.py

@@ -39,4 +39,4 @@
 from .augmentations.trim import Trim
 from .core.composition import Compose, OneOf, SomeOf
 
-__version__ = "0.42.0"
+__version__ = "0.43.0"

docs/changelog.md

@@ -5,6 +5,24 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.43.0] - 2025-09-09
+
+### Changed
+
+* Make `LoudnessNormalization` **400% faster** by switching from pyloudnorm to the new and faster alternative [loudness](https://github.com/iver56/loudness) 
+* Improve `AddBackgroundNoise` performance: Only load/decode the part of the noise file that is needed, instead of the entire file.
+* Bump min Python version to 3.10
+* Switch to a more modern type hint style
+
+### Deprecated
+
+* Deprecate `"pydub"` backend in `Mp3Compression`, because pydub isn't maintained anymore, and depends on audioop, which got removed in Python 3.13. The recommended alternative is `"fast-mp3-augment"`
+
+### Removed
+
+* Remove upper Python version limit to avoid hindering early adopters
+* Remove LRU cache in `AddBackgroundNoise`
+
 ## [0.42.0] - 2025-07-04
 
 ### Added
@@ -682,6 +700,8 @@ Thanks to karpnv
 
 * Initial release. Includes only one transform: `AddGaussianNoise`
 
+[0.43.0]: https://github.com/iver56/audiomentations/compare/v0.42.0...v0.43.0
+
 [0.42.0]: https://github.com/iver56/audiomentations/compare/v0.41.0...v0.42.0
 
 [0.41.0]: https://github.com/iver56/audiomentations/compare/v0.40.0...v0.41.0

docs/waveform_transforms/loudness_normalization.md

@@ -1,6 +1,6 @@
 # `LoudnessNormalization`
 
-_Added in v0.14.0_
+_Added in v0.14.0, updated in v0.43.0_
 
 Apply a constant amount of gain to match a specific loudness (in LUFS). This is an
 implementation of ITU-R BS.1770-4.

docs/waveform_transforms/mp3_compression.md

@@ -1,6 +1,6 @@
 # `Mp3Compression`
 
-_Added in v0.12.0, updated in v0.42.0_
+_Added in v0.12.0, updated in v0.43.0_
 
 Compress the audio using an MP3 encoder to lower the audio quality. This may help machine
 learning models deal with compressed, low-quality audio.