Mikkeli Matlock
d7782bb9d9
Profiled hot spots on a 4-min track and cut the worst offenders: - Remove BPM: librosa.beat.beat_track ran on every load (~3.7s) for a number no better than tapping by hand. Dropped from AudioFile + the metadata panel. - LUFS short-term: replace 474 per-window pyloudnorm.integrated_loudness calls with one K-weighting pass (reusing pyloudnorm's own filter coefficients) + a vectorised sliding mean-square. This is true *ungated* EBU R128 short-term (the old loop wrongly gated each 3s window). Integrated + LRA still use pyloudnorm's gated calls. ~3.8s -> ~1.9s. - PSR: reuse LUFS's short-term series (memoised on the AudioFile) + vectorised sample-peak. ~3.0s -> ~0.2s. - True Peak: oversample the whole signal once, then an O(N) running max over windows instead of per-window resample_poly. Bit-identical to the old loop (max|diff| 0.0000 dB). ~2.1s -> ~1.1s. - Crest Factor: peaks via the same O(N) running max (last per-window loop gone). lufs+psr+true_peak: ~9.2s -> ~3.2s, plus ~3.7s of BPM removed from every load. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
63 lines
2.1 KiB
Python
63 lines
2.1 KiB
Python
import os
|
|
|
|
import librosa
|
|
import numpy as np
|
|
|
|
from mutagen.mp3 import MP3
|
|
from mutagen.easyid3 import EasyID3
|
|
|
|
from font_manager import safe_title
|
|
|
|
|
|
def _try_mp3_tags(file_path):
|
|
try:
|
|
return MP3(file_path, ID3=EasyID3)
|
|
except Exception:
|
|
return None
|
|
|
|
|
|
class AudioFile:
|
|
def __init__(self, file_path):
|
|
self.file_path = file_path
|
|
audio = _try_mp3_tags(self.file_path)
|
|
artist = audio.get('artist', [None])[0] if audio is not None else None
|
|
title = audio.get('title', [None])[0] if audio is not None else None
|
|
if artist and title:
|
|
self.song_name = safe_title(f"{artist} - {title}")
|
|
else:
|
|
self.song_name = safe_title(os.path.basename(self.file_path))
|
|
|
|
# librosa.load normalises to [-1.0, 1.0]. sr=None preserves the file's
|
|
# native sample rate; without it librosa resamples to 22050 Hz, which would
|
|
# discard everything above ~11 kHz (the entire top octave) before analysis.
|
|
self.y, self.sr = librosa.load(file_path, sr=None)
|
|
self.y_mono = librosa.to_mono(self.y)
|
|
self.max_amplitude = np.max(np.abs(self.y_mono))
|
|
self.avg_amplitude = np.mean(np.abs(self.y_mono))
|
|
# BPM intentionally not computed: librosa.beat.beat_track cost ~3.7s on a
|
|
# 4-min track for a number that's no better than tapping it by hand.
|
|
|
|
def get_energy_levels_over_time(self, window=10, hop=2):
|
|
"""Compute rolling RMS power.
|
|
|
|
Args:
|
|
window: Rolling window length in seconds.
|
|
hop: Hop length in seconds.
|
|
"""
|
|
if (not hasattr(self, 'window')) or (self.window != window) or (self.hop != hop):
|
|
self.window, self.hop = window, hop
|
|
if not hasattr(self, 'rms_array'):
|
|
window_samples = window * self.sr
|
|
hop_samples = hop * self.sr
|
|
self.rms_array = librosa.feature.rms(
|
|
y=self.y, frame_length=window_samples, hop_length=hop_samples
|
|
)
|
|
|
|
def get_times(self):
|
|
"""Time-axis values matching the RMS frames."""
|
|
if not hasattr(self, 'rms_array'):
|
|
self.get_energy_levels_over_time()
|
|
return librosa.frames_to_time(
|
|
np.arange(self.rms_array.shape[1]), sr=self.sr, hop_length=self.hop * self.sr
|
|
)
|