Audalign Other Functions

Methods and configuration

BaseConfig

fields for:

• freq_threshold: used for every technique
• multiprocessing: used throughout audalign
• num_processors: used throughout audalign. If None, uses all system processes.

FingerprintConfig

getters and setters for:

• hash_style: "base", "panako", "panako_mod", and "base_three"
• accuracy: 1 - 4

FingerprintRecognizer Methods

• save_fingerprinted_files: saves fingerprints to file in either json or pickle.
• load_fingerprinted_files: loads fingerprints from either json or pickle file.
• filter_duplicates: used internally, shouldn't use this unless manually messing with fingerprints.
• clear_fingerprints: Clears all fingerprints.
• fingerprint_directory: fingerprints directory for use in recognize.
• fingerprint_file: fingerprints file for use in recognize.

Helpful and Useful Functions in Audalign

• rank_alignment: ranks alignment or recognition. Each recognition is given a value from 1-10. Not definitive, but a good indicator.
• write_processed_file: writes audio file to another audio file after regular processing. 1 channel, normalized, 16 bit, 44100 hertz.
• plot: plots peaks calculated on spectrogram of audio file.
• get_metadata: returns metadata returned from ffmpeg.
• write_shifted_file: like write_processed_file but places given seconds of silence in front.
• convert_audio_file: converts audio file to given format if ffmpeg supports it.
• remove_noise_file: wrap around timsainburg's Noisereduce
• remove_noise_directory: wrap around timsainburg's Noisereduce
• uniform_level_file: levels the file using windows that are either normalized or leveled by average sound level
• uniform_level_directory: levels the directory using windows that are either normalized or leveled by average sound level
• recalc_shifts: recalculates shifts from previous alignment. option to choose key to calculate by or match index.
• write_shifts_from_results: writes results from previous alignment. can choose different different source files.
  • unprocessed parameter writes files without normalized or converting sample rate or sample width
• pretty_print_results: nifty pretty printer for recognitions or alignments. Internally calls the other two pretty prints
• pretty_print_recognition: nifty pretty printer for recognitions.
• pretty_print_alignment: nifty pretty printer for alignments.

Uniform leveling pairs well with removing noise. Leveling the file boosts important sound events without losing any data so they have a larger influence in alignments. Removing the noise afterward ensures boosted noise isn't reflected in alignments as much.