### Install Audiomentations

Source: https://github.com/iver56/audiomentations/blob/main/README.md

Install the library using pip. This command is compatible with Linux, macOS, and Windows.

```bash
pip install audiomentations
```

--------------------------------

### Install Audiomentations

Source: https://context7.com/iver56/audiomentations/llms.txt

Install the audiomentations library using pip. Install optional extras for additional transforms.

```bash
pip install audiomentations

# Install optional extras (Limiter, LoudnessNormalization, Mp3Compression, RoomSimulator)
pip install audiomentations[extras]
```

--------------------------------

### Limiter Usage Examples

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/limiter.md

Examples demonstrating how to use the Limiter transform with different threshold modes.

```APIDOC
## Limiter Transform Examples

### Threshold relative to signal peak

```python
from audiomentations import Limiter

transform = Limiter(
    min_threshold_db=-16.0,
    max_threshold_db=-6.0,
    threshold_mode="relative_to_signal_peak",
    p=1.0,
)

augmented_sound = transform(my_waveform_ndarray, sample_rate=16000)
```

### Absolute threshold

```python
from audiomentations import Limiter

transform = Limiter(
    min_threshold_db=-16.0,
    max_threshold_db=-6.0,
    threshold_mode="absolute",
    p=1.0,
)

augmented_sound = transform(my_waveform_ndarray, sample_rate=16000)
```

## Limiter API Parameters

- **min_threshold_db** (float): Minimum threshold in Decibel. Default: -24.0.
- **max_threshold_db** (float): Maximum threshold in Decibel. Default: -2.0.
- **min_attack** (float): Minimum attack time in seconds. Default: 0.0005.
- **max_attack** (float): Maximum attack time in seconds. Default: 0.025.
- **min_release** (float): Minimum release time in seconds. Default: 0.05.
- **max_release** (float): Maximum release time in seconds. Default: 0.7.
- **threshold_mode** (str): Specifies the mode for determining the threshold. Choices: `"relative_to_signal_peak"`, `"absolute"`. Default: `"relative_to_signal_peak"`.
- **p** (float): The probability of applying this transform. Range: [0.0, 1.0]. Default: 0.5.
```

--------------------------------

### Usage Example

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/pitch_shift.md

Demonstrates how to instantiate and use the PitchShift augmentation with a waveform and sample rate.

```APIDOC
```python
from audiomentations import PitchShift

transform = PitchShift(
    min_semitones=-5.0,
    max_semitones=5.0,
    p=1.0
)

augmented_sound = transform(my_waveform_ndarray, sample_rate=44100)
```
```

--------------------------------

### Usage Example

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/air_absorption.md

Demonstrates how to instantiate and use the AirAbsorption transform with custom parameters.

```APIDOC
```python
from audiomentations import AirAbsorption

transform = AirAbsorption(
    min_distance=10.0,
    max_distance=50.0,
    p=1.0,
)

augmented_sound = transform(my_waveform_ndarray, sample_rate=48000)
```
```

--------------------------------

### Build Distribution Files

Source: https://github.com/iver56/audiomentations/blob/main/packaging.md

Run these commands to create source distribution (sdist) and wheel binary distribution files. Ensure you have setuptools and wheel installed.

```bash
python setup.py sdist bdist_wheel
```

--------------------------------

### TanhDistortion Usage

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/tanh_distortion.md

Example of how to instantiate and use the TanhDistortion transform with custom parameters.

```APIDOC
## TanhDistortion API

This section details the parameters available for the `TanhDistortion` transform.

### Parameters

*   **min_distortion** (`float`): Range: [0.0, 1.0]. Default: `0.01`. Minimum "amount" of distortion to apply to the signal.
*   **max_distortion** (`float`): Range: [0.0, 1.0]. Default: `0.7`. Maximum "amount" of distortion to apply to the signal.
*   **p** (`float`): Range: [0.0, 1.0]. Default: `0.5`. The probability of applying this transform.

### Usage Example

```python
from audiomentations import TanhDistortion

transform = TanhDistortion(
    min_distortion=0.01,
    max_distortion=0.7,
    p=1.0
)

augmented_sound = transform(my_waveform_ndarray, sample_rate=16000)
```
```

--------------------------------

### Basic Reverse Augmentation Usage

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/reverse.md

This example demonstrates how to apply the Reverse augmentation to an audio waveform with a probability of 1.0.

```python
from audiomentations import Reverse

transform = Reverse(p=1.0)

augmented_sound = transform(my_waveform_ndarray, sample_rate=44100)
```

--------------------------------

### Upload to PyPI

Source: https://github.com/iver56/audiomentations/blob/main/packaging.md

Use twine to upload the distribution files to the Python Package Index. Make sure you have twine installed and are authenticated with PyPI.

```bash
python -m twine upload dist/*
```

--------------------------------

### PitchShift Usage Example

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/pitch_shift.md

Instantiate and apply the PitchShift transform to an audio waveform. Ensure the waveform and sample rate are provided.

```python
from audiomentations import PitchShift

transform = PitchShift(
    min_semitones=-5.0,
    max_semitones=5.0,
    p=1.0
)

augmented_sound = transform(my_waveform_ndarray, sample_rate=44100)
```

--------------------------------

### BandPassFilter Usage Example

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/band_pass_filter.md

Instantiate and apply the BandPassFilter to an audio waveform. Ensure the input is a NumPy array and the sample rate is provided.

```python
from audiomentations import BandPassFilter

transform = BandPassFilter(min_center_freq=100.0, max_center_freq=6000, p=1.0)

augmented_sound = transform(my_waveform_ndarray, sample_rate=48000)
```

--------------------------------

### Apply Post Gain to Match Input Loudness

Source: https://context7.com/iver56/audiomentations/llms.txt

Wraps another transform and applies a compensating gain after it runs to match the output loudness (RMS, LUFS, or peak) to the input loudness. This example applies PitchShift and then adjusts gain.

```python
from audiomentations import PostGain, PitchShift
import numpy as np

transform = PostGain(
    transform=PitchShift(min_semitones=-4, max_semitones=4, p=1.0),
    method="same_rms",  # "same_rms", "same_lufs", "peak_normalize_always",
                        # or "peak_normalize_if_too_loud"
)

samples = np.random.uniform(-0.2, 0.2, size=(44100,)).astype(np.float32)
augmented = transform(samples=samples, sample_rate=44100)
# PitchShift applied, then output gained so RMS matches input RMS
```

--------------------------------

### Aliasing Transform Usage

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/aliasing.md

Example of how to use the Aliasing transform with specified sample rate ranges and probability.

```APIDOC
## Aliasing Transform

### Description

Downsample the audio to a lower sample rate by linear interpolation, without low-pass filtering it first, resulting in aliasing artifacts. After the downsampling, the signal gets upsampled to the original sample rate again, so the length of the output becomes the same as the length of the input.

### Parameters

#### Initialization Parameters

- **min_sample_rate** (int) - Required - Minimum target sample rate to downsample to. Unit: Hz. Range: [2, ∞).
- **max_sample_rate** (int) - Required - Maximum target sample rate to downsample to. Unit: Hz. Range: [2, ∞).
- **p** (float) - Optional - Default: `0.5`. The probability of applying this transform. Range: [0.0, 1.0].

### Usage Example

```python
from audiomentations import Aliasing

transform = Aliasing(min_sample_rate=8000, max_sample_rate=30000, p=1.0)

augmented_sound = transform(my_waveform_ndarray, sample_rate=44100)
```
```

--------------------------------

### Compose Audio Augmentations

Source: https://github.com/iver56/audiomentations/blob/main/docs/index.md

Use Compose to apply a sequence of audio transforms. This example demonstrates combining Gaussian noise, time stretching, pitch shifting, and sample shifting. Ensure input samples are float32 between -1 and 1.

```python
from audiomentations import Compose, AddGaussianNoise, TimeStretch, PitchShift, Shift
import numpy as np

augment = Compose([
    AddGaussianNoise(min_amplitude=0.001, max_amplitude=0.015, p=0.5),
    TimeStretch(min_rate=0.8, max_rate=1.25, p=0.5),
    PitchShift(min_semitones=-4, max_semitones=4, p=0.5),
    Shift(p=0.5),
])

# Generate 2 seconds of dummy audio for the sake of example
samples = np.random.uniform(low=-0.2, high=0.2, size=(32000,)).astype(np.float32)

# Augment/transform/perturb the audio data
augmented_samples = augment(samples=samples, sample_rate=16000)
```

--------------------------------

### Load Audio with Librosa and Soundfile

Source: https://github.com/iver56/audiomentations/blob/main/docs/guides/multichannel_audio_array_shapes.md

Demonstrates loading audio files using librosa (mono and multichannel) and soundfile. Note the different output shapes for stereo audio.

```python
import librosa
import soundfile as sf

# Librosa, mono
y, sr = librosa.load("stereo_audio_example.wav", sr=None, mono=True)
print(y.shape)  # (117833,)

# Librosa, multichannel
y, sr = librosa.load("stereo_audio_example.wav", sr=None, mono=False)
print(y.shape)  # (2, 117833)

# Soundfile
y, sr = sf.read("stereo_audio_example.wav")
print(y.shape)  # (117833, 2)
```

--------------------------------

### Initialize Mp3Compression Transform

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/mp3_compression.md

Instantiate the Mp3Compression transform with specified bitrate range, backend, and probability. The 'fast-mp3-augment' backend is recommended for performance. Ensure the sample rate is supported by the chosen backend.

```python
from audiomentations import Mp3Compression

transform = Mp3Compression(
    min_bitrate=16,
    max_bitrate=96,
    backend="fast-mp3-augment",
    preserve_delay=False,
    p=1.0
)
```

--------------------------------

### AddGaussianSNR Usage Example

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/add_gaussian_snr.md

Instantiate AddGaussianSNR with desired SNR range and probability, then apply it to a waveform.

```python
from audiomentations import AddGaussianSNR

transform = AddGaussianSNR(
    min_snr_db=5.0,
    max_snr_db=40.0,
    p=1.0
)

augmented_sound = transform(my_waveform_ndarray, sample_rate=16000)
```

--------------------------------

### Initialize and Apply TimeStretch

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/time_stretch.md

Instantiate TimeStretch with desired parameters for rate limits and length preservation, then apply it to an audio waveform.

```python
from audiomentations import TimeStretch

transform = TimeStretch(
    min_rate=0.8,
    max_rate=1.25,
    leave_length_unchanged=True,
    p=1.0
)

augmented_sound = transform(my_waveform_ndarray, sample_rate=16000)
```

--------------------------------

### RoomSimulator: Simulate Parametric Room Acoustics

Source: https://context7.com/iver56/audiomentations/llms.txt

Synthesizes a room impulse response procedurally and convolves it with audio. Allows fine-grained control over room dimensions, absorption, and positions. The output length may differ from the input. Ensure `numpy` is imported.

```python
from audiomentations import RoomSimulator
import numpy as np

transform = RoomSimulator(
    min_size_x=3.6, max_size_x=5.6,
    min_size_y=3.6, max_size_y=3.9,
    min_size_z=2.4, max_size_z=3.0,
    min_absorption_value=0.075, max_absorption_value=0.4,
    calculation_mode="rt60",   # or "absorption"
    min_target_rt60=0.15,      # recording studio ~0.3 s, concert hall ~1.5 s
    max_target_rt60=0.8,
    use_ray_tracing=True,
    max_order=3,
    leave_length_unchanged=False,
    p=1.0,
)

samples = np.random.uniform(-0.2, 0.2, size=(16000,)).astype(np.float32)
augmented = transform(samples=samples, sample_rate=16000)
# Reverb tail added; output may be longer than input
```

--------------------------------

### RoomSimulator Parameters

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/room_simulator.md

This section details the configurable parameters for the RoomSimulator, including room dimensions, microphone settings, reverberation modes, and simulation options.

```APIDOC
## RoomSimulator API

This API allows for the simulation of acoustic environments. Below are the configurable parameters:

### Parameters

- **`max_mic_elevation`** (`float`): Maximum elevation of the microphone relative to the source, in radians. Default: `math.pi`.

- **`calculation_mode`** (`str`): Determines how the room's acoustic properties are calculated. Choices: `"rt60"`, `"absorption"`. Default: `"absorption"`.
  - If `"absorption"`, room surfaces are based on `min_absorption_value` and `max_absorption_value`.
  - If `"rt60"`, surface materials are assigned to achieve target RT60 values specified by `min_target_rt60` and `max_target_rt60`.

- **`use_ray_tracing`** (`bool`): Whether to use ray tracing for simulation. Default: `True`. Disabling this improves speed at the cost of accuracy.

- **`max_order`** (`int`): Maximum order of reflections for the Image Source Model. Default: `1`. Higher values add more diffuse reverberation but can slow down the process, especially when `calculation_mode="rt60"`. Recommended values for `"rt60"` are around `3-4`.

- **`leave_length_unchanged`** (`bool`): If `True`, the output audio length will be equal to the input audio length by chopping off the tail. Default: `False`.

- **`padding`** (`float`): Minimum distance in meters between sources/microphones and room walls, floor, or ceiling. Default: `0.1`.

- **`p`** (`float`): The probability of applying this transform. Range: `[0.0, 1.0]`. Default: `0.5`.

- **`ray_tracing_options`** (`Optional[Dict]`): Options for the ray tracer. See `set_ray_tracing` in the pyroomacoustics library for details. Default: `None`.
```

--------------------------------

### Trim Silence with Default Parameters

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/trim.md

Apply the Trim transform to remove silence from the start and end of an audio waveform. Uses the default top_db parameter value.

```python
from audiomentations import Trim

transform = Trim(
    top_db=30.0,
    p=1.0
)

augmented_sound = transform(my_waveform_ndarray, sample_rate=16000)
```

--------------------------------

### Inspect and Freeze Transform Parameters

Source: https://context7.com/iver56/audiomentations/llms.txt

Demonstrates how to inspect the parameters of a transform after it has been called and how to freeze these parameters to apply the exact same augmentation to multiple signals.

```python
from audiomentations import Gain
import numpy as np

augment = Gain(p=1.0)

signal = np.random.uniform(-0.2, 0.2, size=(32000,)).astype(np.float32)
target = np.random.uniform(-0.2, 0.2, size=(32000,)).astype(np.float32)

augmented_signal = augment(samples=signal, sample_rate=16000)
print(augment.parameters)
# {'should_apply': True, 'amplitude_ratio': 0.9688148624484364}

augment.freeze_parameters()  # lock parameters at current values
augmented_target = augment(samples=target, sample_rate=16000)
print(augment.parameters)
# {'should_apply': True, 'amplitude_ratio': 0.9688148624484364}  (identical)

augment.unfreeze_parameters()  # allow new random parameters next call
```

--------------------------------

### Add Background Noise with Absolute RMS

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/add_background_noise.md

Use this when you want to control the absolute loudness of the background noise, irrespective of the input signal's level. Includes an example of applying a noise transform.

```python
from audiomentations import AddBackgroundNoise, PolarityInversion

transform = AddBackgroundNoise(
    sounds_path="/path/to/folder_with_sound_files",
    noise_rms="absolute",
    min_absolute_rms_db=-45.0,
    max_absolute_rms_db=-15.0,
    noise_transform=PolarityInversion(),
    p=1.0
)

augmented_sound = transform(my_waveform_ndarray, sample_rate=16000)
```

--------------------------------

### Add Background Noise with Relative RMS

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/add_background_noise.md

Use this when you want the noise level to be relative to the input signal's level. If the input is silent, no noise will be added. Includes an example of applying a noise transform.

```python
from audiomentations import AddBackgroundNoise, PolarityInversion

transform = AddBackgroundNoise(
    sounds_path="/path/to/folder_with_sound_files",
    min_snr_db=3.0,
    max_snr_db=30.0,
    noise_transform=PolarityInversion(),
    p=1.0
)

augmented_sound = transform(my_waveform_ndarray, sample_rate=16000)
```

--------------------------------

### RoomSimulator

Source: https://context7.com/iver56/audiomentations/llms.txt

Synthesizes a room impulse response procedurally and convolves it with the audio.

```APIDOC
## RoomSimulator

### Description
Synthesizes a room impulse response procedurally using `pyroomacoustics` and convolves it with the audio. Allows fine-grained control over room dimensions, absorption coefficients, source and microphone positions.

### Parameters
- `min_size_x`, `max_size_x` (float) - Minimum and maximum size of the room along the X-axis.
- `min_size_y`, `max_size_y` (float) - Minimum and maximum size of the room along the Y-axis.
- `min_size_z`, `max_size_z` (float) - Minimum and maximum size of the room along the Z-axis.
- `min_absorption_value`, `max_absorption_value` (float) - Minimum and maximum absorption values for the room surfaces.
- `calculation_mode` (str) - Mode for calculation ('rt60' or 'absorption').
- `min_target_rt60`, `max_target_rt60` (float) - Minimum and maximum target reverberation times (RT60).
- `use_ray_tracing` (bool) - Whether to use ray tracing for simulation.
- `max_order` (int) - Maximum order for the simulation.
- `leave_length_unchanged` (bool) - If True, the reverb tail is not added, preserving input length.
- `p` (float) - Probability of applying the transform.
```

--------------------------------

### OneOf Transform Selection

Source: https://github.com/iver56/audiomentations/blob/main/docs/index.md

Use OneOf to randomly select and apply one transform from a list. Weights can be provided to influence the probability of selection for each transform. This example uses different pitch shifting methods.

```python
from audiomentations import OneOf, PitchShift

pitch_shift = OneOf(
    transforms=[
        PitchShift(method="librosa_phase_vocoder"),
        PitchShift(method="signalsmith_stretch"),
    ],
    p=1.0,
    weights=[0.1, 0.9],
)
```

--------------------------------

### Padding Augmentation

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/padding.md

This augmentation applies padding to the audio signal. You can control the mode of padding, the minimum and maximum fraction of the signal to pad, and which section (start or end) to pad. The probability of applying the transform can also be set.

```APIDOC
## Padding Augmentation

Applies padding to the audio signal by taking a fraction of the start or end and replacing that portion with padding. This can be useful for training ML models on padded inputs.

### Parameters

- **mode** (str) - Optional - Default: "silence". Choices: "silence", "wrap", "reflect". Padding mode.
- **min_fraction** (float) - Optional - Default: 0.01. Minimum fraction of the signal duration to be padded. Range: [0.0, 1.0].
- **max_fraction** (float) - Optional - Default: 0.7. Maximum fraction of the signal duration to be padded. Range: [0.0, 1.0].
- **pad_section** (str) - Optional - Default: "end". Choices: "start", "end". Which part of the signal should be replaced with padding.
- **p** (float) - Optional - Default: 0.5. The probability of applying this transform. Range: [0.0, 1.0].
```

--------------------------------

### PostGain Transform

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/post_gain.md

The PostGain transform can be initialized with a callable transform and a method for gain adjustment. The available methods are 'same_rms', 'same_lufs', 'peak_normalize_always', and 'peak_normalize_if_too_loud'.

```APIDOC
## PostGain

### Description

Gain up or down the audio after the given transform (or set of transforms) has processed the audio. There are several methods that determine how the audio should be gained. `PostGain` can be useful for compensating for any gain differences introduced by a (set of) transform(s), or for preventing clipping in the output.

### Parameters

- **`transform`** (`Callable[[NDArray[np.float32], int], NDArray[np.float32]]`) - Required - A callable to be applied. It should input samples (ndarray), sample_rate (int) and optionally some user-defined keyword arguments.
- **`method`** (`str`) - Required - This parameter defines the method for choosing the post gain amount. Choices include: `"same_rms"`, `"same_lufs"`, `"peak_normalize_always"`, or `"peak_normalize_if_too_loud"`.

### Method Details

- **`"same_rms"`**: The sound gets post-gained so that the RMS (Root Mean Square) of the output matches the RMS of the input.
- **`"same_lufs"`**: The sound gets post-gained so that the LUFS (Loudness Units Full Scale) of the output matches the LUFS of the input.
- **`"peak_normalize_always"`**: The sound gets peak normalized (gained up or down so that the absolute value of the most extreme sample in the output is 1.0).
- **`"peak_normalize_if_too_loud"`**: The sound gets peak normalized if it is too loud (max absolute value greater than 1.0). This option can be useful for avoiding clipping.
```

--------------------------------

### BandPassFilter Initialization and Usage

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/band_pass_filter.md

This snippet shows how to initialize the BandPassFilter with custom parameters and apply it to an audio waveform.

```APIDOC
## BandPassFilter

Apply band-pass filtering to the input audio. Filter steepness (6/12/18... dB / octave) is parametrized. Can also be set for zero-phase filtering (will result in a 6 dB drop at cutoffs).

### Parameters

- **min_center_freq** (float): Minimum center frequency in hertz. Default: `200.0`.
- **max_center_freq** (float): Maximum center frequency in hertz. Default: `4000.0`.
- **min_bandwidth_fraction** (float): Minimum bandwidth relative to center frequency. Range: [0.0, 2.0]. Default: `0.5`.
- **max_bandwidth_fraction** (float): Maximum bandwidth relative to center frequency. Range: [0.0, 2.0]. Default: `1.99`.
- **min_rolloff** (int): Minimum filter roll-off (in dB/octave). Must be a multiple of 6 (or 12 if `zero_phase` is `True`). Default: `12`.
- **max_rolloff** (int): Maximum filter roll-off (in dB/octave). Must be a multiple of 6 (or 12 if `zero_phase` is `True`). Default: `24`.
- **zero_phase** (bool): Whether filtering should be zero phase. Default: `False`.
- **p** (float): The probability of applying this transform. Range: [0.0, 1.0]. Default: `0.5`.

### Usage Example

```python
from audiomentations import BandPassFilter

transform = BandPassFilter(min_center_freq=100.0, max_center_freq=6000, p=1.0)

augmented_sound = transform(my_waveform_ndarray, sample_rate=48000)
```
```

--------------------------------

### Apply Custom Callable with Lambda

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/lambda.md

Use the Lambda transform to apply a custom Python function to audio samples. The callable should accept samples and sample rate, and can optionally take additional keyword arguments. This example demonstrates integrating a custom gain adjustment for the left channel within a `OneOf` composition.

```python
import random

from audiomentations import Lambda, OneOf, Gain


def gain_only_left_channel(samples, sample_rate):
    samples[0, :] *= random.uniform(0.8, 1.25)
    return samples


transform = OneOf(
    transforms=[Lambda(transform=gain_only_left_channel, p=1.0), Gain(p=1.0)]
)

augmented_sound = transform(my_stereo_waveform_ndarray, sample_rate=16000)
```

--------------------------------

### AdjustDuration: Trim or Pad to Fixed Length

Source: https://context7.com/iver56/audiomentations/llms.txt

Trims or pads audio to a target duration, specified in samples or seconds. Useful for batching variable-length audio. The `padding_mode` can be "silence", "wrap", or "reflect", and `padding_position` can be "start" or "end". Ensure `numpy` is imported.

```python
from audiomentations import AdjustDuration
import numpy as np

# Target specified in samples
transform_samples = AdjustDuration(duration_samples=24000, p=1.0)

# Target specified in seconds
transform_seconds = AdjustDuration(
    duration_seconds=3.0,
    padding_mode="silence",   # "silence", "wrap", or "reflect"
    padding_position="end",   # "start" or "end"
    p=1.0,
)

samples = np.random.uniform(-0.2, 0.2, size=(40000,)).astype(np.float32)
augmented = transform_samples(samples=samples, sample_rate=16000)
assert augmented.shape == (24000,)  # exactly 1.5 s at 16 kHz
```

--------------------------------

### Convert Channels-Last to Channels-First with NumPy

Source: https://github.com/iver56/audiomentations/blob/main/docs/guides/multichannel_audio_array_shapes.md

Shows how to convert audio data from the channels-last format (samples, channels) to the channels-first format (channels, samples) using NumPy's transpose operation.

```python
import numpy as np

# Assuming y is your audio data in channels-last format
y_transposed = np.transpose(y)

# Alternative, shorter syntax:
y_transposed = y.T
```

--------------------------------

### Apply BitCrush Transform in Python

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/bit_crush.md

Instantiate the BitCrush transform with specified minimum and maximum bit depths and apply it to an audio waveform. Set p=1.0 to ensure the transform is always applied.

```python
from audiomentations import BitCrush

transform = BitCrush(min_bit_depth=5, max_bit_depth=14, p=1.0)

augmented_sound = transform(my_waveform_ndarray, sample_rate=16000)
```

--------------------------------

### Apply Impulse Response Transform

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/apply_impulse_response.md

Instantiate the ApplyImpulseResponse transform and apply it to a waveform. Ensure the ir_path points to a folder containing impulse response audio files. Set p=1.0 to guarantee application.

```python
from audiomentations import ApplyImpulseResponse

transform = ApplyImpulseResponse(ir_path="/path/to/sound_folder", p=1.0)

augmented_sound = transform(my_waveform_ndarray, sample_rate=48000)
```

--------------------------------

### Apply Mp3Compression to Waveform

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/mp3_compression.md

Apply the initialized Mp3Compression transform to a waveform numpy array. The sample rate must be provided and should be one of the supported rates for the selected backend.

```python
augmented_sound = transform(my_waveform_ndarray, sample_rate=48000)
```

--------------------------------

### Initialize and Apply AirAbsorption

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/air_absorption.md

Instantiate the AirAbsorption transform with specified distance ranges and apply it to a waveform. Ensure the sample rate is appropriate for high-frequency content.

```python
from audiomentations import AirAbsorption

transform = AirAbsorption(
    min_distance=10.0,
    max_distance=50.0,
    p=1.0,
)

augmented_sound = transform(my_waveform_ndarray, sample_rate=48000)
```

--------------------------------

### Ensure C-Contiguous Array for Audio Processing

Source: https://github.com/iver56/audiomentations/blob/main/docs/guides/multichannel_audio_array_shapes.md

Explains the importance of C-contiguous memory layout for some audio processing libraries and demonstrates how to ensure it using `np.ascontiguousarray` after transposing.

```python
y_transposed_contiguous = np.ascontiguousarray(y_transposed)
```

--------------------------------

### Inspect Parameters in Compose Pipeline

Source: https://context7.com/iver56/audiomentations/llms.txt

Shows how to iterate through transforms within a Compose pipeline and inspect the parameters of each individual transform after augmentation.

```python
from audiomentations import Compose, AddGaussianNoise, TimeStretch, PitchShift, Shift
import numpy as np

augment = Compose([
    AddGaussianNoise(p=0.5),
    TimeStretch(p=0.5),
    PitchShift(p=0.5),
    Shift(p=0.5),
])
samples = np.random.uniform(-0.2, 0.2, size=(32000,)).astype(np.float32)
augmented = augment(samples=samples, sample_rate=16000)

for t in augment.transforms:
    print(f"{t.__class__.__name__}: {t.parameters}")
# AddGaussianNoise: {'should_apply': True, 'amplitude': 0.0027}
# TimeStretch: {'should_apply': True, 'rate': 1.158}
# PitchShift: {'should_apply': False}
# Shift: {'should_apply': False}
```

--------------------------------

### Source and Microphone Placement

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/room_simulator.md

Defines the ranges for placing sound sources and microphones within the simulated room.

```APIDOC
## Source and Microphone Placement

### Parameters

- **min_source_x** (float) - Optional - Minimum x location of the source in meters. Default: `0.1`.
- **max_source_x** (float) - Optional - Maximum x location of the source in meters. Default: `3.5`.
- **min_source_y** (float) - Optional - Minimum y location of the source in meters. Default: `0.1`.
- **max_source_y** (float) - Optional - Maximum y location of the source in meters. Default: `2.7`.
- **min_source_z** (float) - Optional - Minimum z location of the source in meters. Default: `1.0`.
- **max_source_z** (float) - Optional - Maximum z location of the source in meters. Default: `2.1`.
- **min_mic_distance** (float) - Optional - Minimum distance of the microphone from the source in meters. Default: `0.15`.
- **max_mic_distance** (float) - Optional - Maximum distance of the microphone from the source in meters. Default: `0.35`.
- **min_mic_azimuth** (float) - Optional - Minimum azimuth (angle around z axis) of the microphone relative to the source in radians. Default: `-math.pi`.
- **max_mic_azimuth** (float) - Optional - Maximum azimuth (angle around z axis) of the microphone relative to the source in radians. Default: `math.pi`.
- **min_mic_elevation** (float) - Optional - Minimum elevation of the microphone relative to the source in radians. Default: `-math.pi`.
```

--------------------------------

### SevenBandParametricEQ

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/seven_band_parametric_eq.md

Initializes the SevenBandParametricEQ transform with configurable parameters for gain, Q values, and center frequencies.

```APIDOC
## SevenBandParametricEQ()

### Description
Initializes the SevenBandParametricEQ transform, which applies a 7-band parametric equalizer to audio signals. This transform can adjust the volume of different frequency bands by applying randomized gains, Q values, and center frequencies.

### Parameters
#### Keyword Arguments
- **min_gain_db** (float) - Optional - Default: -12.0. The minimum number of dB to cut or boost a band.
- **max_gain_db** (float) - Optional - Default: 12.0. The maximum number of dB to cut or boost a band.
- **p** (float) - Optional - Default: 0.5. The probability of applying this transform. Must be in the range [0.0, 1.0].
```

--------------------------------

### BandStopFilter Parameters

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/band_stop_filter.md

Configuration options for the BandStopFilter augmentation.

```APIDOC
## BandStopFilter API

### Parameters

- **min_center_freq** (float, optional): Minimum center frequency in hertz. Default: `200.0`.
- **max_center_freq** (float, optional): Maximum center frequency in hertz. Default: `4000.0`.
- **min_bandwidth_fraction** (float, optional): Minimum bandwidth fraction relative to center frequency. Default: `0.5`.
- **max_bandwidth_fraction** (float, optional): Maximum bandwidth fraction relative to center frequency. Default: `1.99`.
- **min_rolloff** (int, optional): Minimum filter roll-off in dB/octave. Must be a multiple of 6 (or 12 if `zero_phase` is `True`). Default: `12`.
- **max_rolloff** (int, optional): Maximum filter roll-off in dB/octave. Must be a multiple of 6 (or 12 if `zero_phase` is `True`). Default: `24`.
- **zero_phase** (bool, optional): Whether filtering should be zero phase. If `True`, it does not affect the phase but sounds 3 dB lower at the cutoff frequency compared to the non-zero phase case (6 dB vs. 3 dB). It is also twice as slow. Default: `False`.
- **p** (float, optional): The probability of applying this transform. Range: [0.0, 1.0]. Default: `0.5`.
```

--------------------------------

### Freezing and Reusing Transform Parameters

Source: https://github.com/iver56/audiomentations/blob/main/docs/guides/transform_parameters.md

To apply the exact same transformation to multiple audio inputs, call `.freeze_parameters()` after the first application. This preserves the chosen parameters. Call `.unfreeze_parameters()` to resume random parameter selection.

```python
from audiomentations import Gain
import numpy as np

augment = Gain(p=1.0)

samples = np.random.uniform(low=-0.2, high=0.2, size=(32000,)).astype(np.float32)
samples2 = np.random.uniform(low=-0.2, high=0.2, size=(32000,)).astype(np.float32)

augmented_samples = augment(samples=samples, sample_rate=16000)
augment.freeze_parameters()
print(augment.parameters)
augmented_samples2 = augment(samples=samples2, sample_rate=16000)
print(augment.parameters)
augment.unfreeze_parameters()
```

--------------------------------

### Trim Transform

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/trim.md

This snippet demonstrates how to apply the Trim transform to an audio waveform. It initializes the transform with a specific `top_db` value and probability `p`, then applies it to a NumPy array representing the waveform.

```APIDOC
## Trim Transform

### Description

Applies the Trim transformation to an audio waveform to remove leading and trailing silence.

### Parameters

#### `top_db` (float)

- Required: No
- Default: `30.0`
- Description: The threshold in Decibels below which audio is considered silence and will be trimmed. A lower value means more silence will be trimmed.

#### `p` (float)

- Required: No
- Default: `0.5`
- Range: [0.0, 1.0]
- Description: The probability of applying this transform.

### Usage Example

```python
from audiomentations import Trim

transform = Trim(
    top_db=30.0,
    p=1.0
)

augmented_sound = transform(samples=my_waveform_ndarray, sample_rate=16000)
```

### Input-Output Example

This example shows the effect of the Trim transform on an audio signal, visualizing the input and transformed waveforms and spectrograms, along with playable audio controls.
```

--------------------------------

### Mp3Compression API

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/mp3_compression.md

Configuration options for the Mp3Compression transform.

```APIDOC
## Mp3Compression

Compress the audio using an MP3 encoder to lower the audio quality. This may help machine learning models deal with compressed, low-quality audio.

### Parameters

*   **`min_bitrate`** (int): Minimum bitrate in kbps. Range: [8, `max_bitrate`]. Default: `8`.
*   **`max_bitrate`** (int): Maximum bitrate in kbps. Range: [`min_bitrate`, 320]. Default: `64`.
*   **`backend`** (str): The backend to use for MP3 compression. Choices: `"fast-mp3-augment"`, `"pydub"`, `"lameenc"`. Default: `"fast-mp3-augment"`.
    *   `"fast-mp3-augment"`: In-memory computation with parallel threads. Uses LAME encoder and minimp3 decoder.
    *   `"pydub"`: Uses pydub + ffmpeg. Does not delay output. Slow, writes temporary files. Deprecated.
    *   `"lameenc"`: Writes temporary files. Introduces delay. Deprecated.
*   **`preserve_delay`** (bool): If `False`, output length and timing match input. If `True`, includes encoder and filter delay, making output longer and delayed. Default: `False`.
*   **`quality`** (int): LAME-specific parameter controlling quality vs. speed. Range: [0, 9]. Default: `7`. Ignored if `backend="pydub"`.
*   **`p`** (float): The probability of applying this transform. Range: [0.0, 1.0]. Default: `0.5`.

### Usage Example

```python
from audiomentations import Mp3Compression

transform = Mp3Compression(
    min_bitrate=16,
    max_bitrate=96,
    backend="fast-mp3-augment",
    preserve_delay=False,
    p=1.0
)

augmented_sound = transform(my_waveform_ndarray, sample_rate=48000)
```
```

--------------------------------

### TimeStretch Transform

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/time_stretch.md

This snippet demonstrates how to instantiate and use the TimeStretch transform. It allows for flexible control over the rate of time stretching, whether the output length should match the input, and the probability of applying the transform.

```APIDOC
## TimeStretch

Change the speed or duration of the signal without changing the pitch.

### Parameters

*   **min_rate** (`float`): Default: `0.8`. Minimum time-stretch rate. Values less than 1.0 slow down the audio.
*   **max_rate** (`float`): Default: `1.25`. Maximum time-stretch rate. Values greater than 1.0 speed up the audio.
*   **leave_length_unchanged** (`bool`): Default: `True`. If `True`, the output audio will have the same duration as the input audio. If `False`, the duration will be altered.
*   **method** (`str`): Default: `"signalsmith_stretch"`. The time stretching algorithm to use. Choices: `"librosa_phase_vocoder"`, `"signalsmith_stretch"`.
*   **p** (`float`): Default: `0.5`. The probability of applying this transform.

### Usage Example

```python
from audiomentations import TimeStretch

transform = TimeStretch(
    min_rate=0.8,
    max_rate=1.25,
    leave_length_unchanged=True,
    p=1.0
)

augmented_sound = transform(my_waveform_ndarray, sample_rate=16000)
```
```

--------------------------------

### Mp3Compression: Simulate MP3 Lossy Compression

Source: https://context7.com/iver56/audiomentations/llms.txt

Encodes audio to MP3 at a random bitrate and decodes it back to simulate lossy compression artifacts. The `backend` can be set to "fast-mp3-augment" (recommended) or others. Ensure `numpy` is imported.

```python
from audiomentations import Mp3Compression
import numpy as np

transform = Mp3Compression(
    min_bitrate=16,
    max_bitrate=96,
    backend="fast-mp3-augment",  # recommended; in-memory, parallel threads
    preserve_delay=False,        # output aligned with input in time
    p=1.0,
)

samples = np.random.uniform(-0.2, 0.2, size=(48000,)).astype(np.float32)
augmented = transform(samples=samples, sample_rate=48000)
# Supported sample rates for fast-mp3-augment: 8000, 11025, 12000, 16000,
# 22050, 24000, 32000, 44100, 48000
```

--------------------------------

### LoudnessNormalization Parameters

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/loudness_normalization.md

Configuration options for the LoudnessNormalization transform.

```APIDOC
## LoudnessNormalization

Apply a constant amount of gain to match a specific loudness (in LUFS). This is an implementation of ITU-R BS.1770-4.

### Parameters

- **min_lufs** (float) - Optional - Minimum loudness target in LUFS. Default: -31.0.
- **max_lufs** (float) - Optional - Maximum loudness target in LUFS. Default: -13.0.
- **p** (float) - Optional - The probability of applying this transform. Range: [0.0, 1.0]. Default: 0.5.

### Warning

This transform can return samples outside the [-1, 1] range, which may lead to clipping or wrap distortion.
```

--------------------------------

### Apply Tanh Distortion

Source: https://github.com/iver56/audiomentations/blob/main/docs/waveform_transforms/tanh_distortion.md

Instantiate TanhDistortion with specified min/max distortion and probability, then apply it to a waveform.

```python
from audiomentations import TanhDistortion

transform = TanhDistortion(
    min_distortion=0.01,
    max_distortion=0.7,
    p=1.0
)

augmented_sound = transform(my_waveform_ndarray, sample_rate=16000)
```