Additional scripts

Notes

Here are some additional functions that I wanted to share.

The results of find_offsets are cached in cached_offset_searches.pickle

Code

find_offsets(within_file, find_file, multiplier, window=2, number=None, max_tries_number=80, to_print_plots=False)

Finds time codes of appearance audio of find_file in within_file using scipy.

Parameters:

Name	Type	Description	Default
within_file	str	path to the audio file to search within	required
find_file	str	path to the audio file to search for	required
multiplier	float	a multiplier used to calculate the prominence of peaks in the correlation signal. The bigger you set it, the less time codes function returns.	required
window	int	the length of find_file audio to use for the correlation (in seconds)	2
number	int	the goal number of offsets to return	None
max_tries_number	int	maximum number of tries to find the number	80
to_print_plots	bool	whether to print plot before proceeding	False

Returns:

Name	Type	Description
list	list	list of time codes

Source code in ffmpeg_python_utils\other.py

@cache_results
def find_offsets(within_file: str, find_file: str, multiplier: float, window: int = 2, number: int = None,
                 max_tries_number: int = 80, to_print_plots=False) -> list:
    """
    Finds time codes of appearance audio of find_file in within_file using scipy.

    Args:
        within_file (str): path to the audio file to search within
        find_file (str): path to the audio file to search for
        multiplier (float): a multiplier used to calculate the prominence of peaks in the correlation signal. The bigger you set it, the less time codes function returns.
        window (int): the length of find_file audio to use for the correlation (in seconds)
        number (int): the goal number of offsets to return
        max_tries_number (int): maximum number of tries to find the number
        to_print_plots (bool): whether to print plot before proceeding

    Returns:
        list: list of time codes
    """

    y_within, sr_within = librosa.load(within_file, sr=None)
    y_find, _ = librosa.load(find_file, sr=sr_within)
    c = signal.correlate(y_within, y_find[:sr_within * window], mode='valid', method='fft')
    if number is not None and number < 1:
        print_info(f'Number of peaks you are looking for is {number}. Returning empty list.', 'red',
                   C_TO_PRINT_PACKAGE_INFO)
        return []

    elif number == 1:
        c = np.argmax(c)
        if to_print_plots:
            plot_offsets(c, find_file)
        peak = round(c / sr_within, 2)
        return [peak]
    else:
        prominence = int(c[np.argmax(c)] * multiplier)
        counter = 0
        while True:
            try:
                peaks, _ = signal.find_peaks(c, prominence=prominence)
                points_of_time = [round(peak / sr_within, 2) for peak in peaks]
                points_of_time = delete_neighbors(points_of_time)
                if to_print_plots:
                    plot_offsets(c, find_file)
                if counter > max_tries_number:
                    if C_TO_PRINT_PACKAGE_INFO: print(
                        f'Max try number reached. Returning the last time codes. {points_of_time}')
                    return points_of_time
                if number and number != len(points_of_time):
                    print_info(
                        f'Try number № {counter}. Looking for peaks. The goal number is {number}, the number we got is {len(points_of_time)}. Prominence: {prominence}',
                        'white', C_TO_PRINT_PACKAGE_INFO)
                    if C_TO_PRINT_PACKAGE_INFO: print_info(f'Time codes we got: {points_of_time}', 'white')
                    if number > len(points_of_time):
                        diff = number - len(points_of_time)
                        prominence *= 0.80 if diff > 5 else 0.95
                    else:
                        prominence *= 1.01
                    counter += 1
                    continue
                print_info(f'Found specified number={number}. Offsets: {points_of_time}', 'white',
                           C_TO_PRINT_PACKAGE_INFO)
                return points_of_time
            except IndexError:
                prominence *= 0.8
                counter += 1

remove_silence_from_audio_file(input_path, output_path, audio_format='wav', min_silence_len=100, silence_thresh=-45, keep_silence=50)

Removes silence from audio file. Results are cached taking in account hash of the input files and kwargs passed.

Parameters:

Name	Type	Description	Default
input_path	str	The path to the audio file.	required
output_path	str	The path to the directory where the edited audio files will be saved.	required
audio_format	str	The format of the audio output.	'wav'
min_silence_len	int	The minimum length of silence to be removed, in milliseconds.	100
silence_thresh	int	The threshold for silence, in decibels.	-45
keep_silence	int	The length of silence to keep at the beginning and end of the audio file, in milliseconds.	50

Returns:

Name	Type	Description
str	str	output_path

Source code in ffmpeg_python_utils\other.py

def remove_silence_from_audio_file(input_path: str, output_path: str, audio_format: str = 'wav',
                                   min_silence_len: int = 100, silence_thresh: int = -45,
                                   keep_silence: int = 50) -> str:
    """
    Removes silence from audio file. Results are cached taking in account hash of the input files and kwargs passed.

    Args:
        input_path (str): The path to the audio file.
        output_path (str): The path to the directory where the edited audio files will be saved.
        audio_format (str, optional): The format of the audio output.
        min_silence_len (int, optional): The minimum length of silence to be removed, in milliseconds.
        silence_thresh (int, optional): The threshold for silence, in decibels.
        keep_silence (int, optional): The length of silence to keep at the beginning and end of the audio file, in milliseconds.

    Returns:
        str: output_path
    """
    sound = AudioSegment.from_file(input_path, format=audio_format)
    audio_chunks = split_on_silence(sound, min_silence_len=min_silence_len, silence_thresh=silence_thresh,
                                    keep_silence=keep_silence)
    combined = AudioSegment.empty()
    for chunk in audio_chunks:
        combined += chunk
    combined.export(output_path, format=audio_format)
    return output_path