mirror of
https://github.com/w-okada/voice-changer.git
synced 2025-02-03 08:43:57 +03:00
d5561c2212
- Add Crepe Full/Tiny (onnx) - remove test connect for local Refactor: - RVC: comment out module importer
130 lines
3.8 KiB
Python
130 lines
3.8 KiB
Python
import numpy as np
|
|
|
|
from voice_changer.RVC.pitchExtractor import onnxcrepe
|
|
|
|
|
|
###############################################################################
|
|
# Pitch thresholding methods
|
|
###############################################################################
|
|
|
|
|
|
class At:
|
|
"""Simple thresholding at a specified probability value"""
|
|
|
|
def __init__(self, value):
|
|
self.value = value
|
|
|
|
def __call__(self, pitch, periodicity):
|
|
# Make a copy to prevent in-place modification
|
|
pitch = pitch.copy()
|
|
|
|
# Threshold
|
|
pitch[periodicity < self.value] = onnxcrepe.UNVOICED
|
|
return pitch
|
|
|
|
|
|
class Hysteresis:
|
|
"""Hysteresis thresholding"""
|
|
|
|
def __init__(self,
|
|
lower_bound=.19,
|
|
upper_bound=.31,
|
|
width=.2,
|
|
stds=1.7,
|
|
return_threshold=False):
|
|
self.lower_bound = lower_bound
|
|
self.upper_bound = upper_bound
|
|
self.width = width
|
|
self.stds = stds
|
|
self.return_threshold = return_threshold
|
|
|
|
def __call__(self, pitch, periodicity):
|
|
|
|
# Perform hysteresis in log-2 space
|
|
pitch = np.log2(pitch).flatten()
|
|
|
|
# Flatten periodicity
|
|
periodicity = periodicity.flatten()
|
|
|
|
# Ignore confidently unvoiced pitch
|
|
pitch[periodicity < self.lower_bound] = onnxcrepe.UNVOICED
|
|
|
|
# Whiten pitch
|
|
mean, std = np.nanmean(pitch), np.nanstd(pitch)
|
|
pitch = (pitch - mean) / std
|
|
|
|
# Require high confidence to make predictions far from the mean
|
|
parabola = self.width * pitch ** 2 - self.width * self.stds ** 2
|
|
threshold = self.lower_bound + np.clip(parabola, 0, 1 - self.lower_bound)
|
|
threshold[np.isnan(threshold)] = self.lower_bound
|
|
|
|
# Apply hysteresis to prevent short, unconfident voiced regions
|
|
i = 0
|
|
while i < len(periodicity) - 1:
|
|
|
|
# Detect unvoiced to voiced transition
|
|
if periodicity[i] < threshold[i] and periodicity[i + 1] > threshold[i + 1]:
|
|
|
|
# Grow region until next unvoiced or end of array
|
|
start, end, keep = i + 1, i + 1, False
|
|
while end < len(periodicity) and periodicity[end] > threshold[end]:
|
|
if periodicity[end] > self.upper_bound:
|
|
keep = True
|
|
end += 1
|
|
|
|
# Force unvoiced if we didn't pass the confidence required by
|
|
# the hysteresis
|
|
if not keep:
|
|
threshold[start:end] = 1
|
|
|
|
i = end
|
|
|
|
else:
|
|
i += 1
|
|
|
|
# Remove pitch with low periodicity
|
|
pitch[periodicity < threshold] = onnxcrepe.UNVOICED
|
|
|
|
# Unwhiten
|
|
pitch = pitch * std + mean
|
|
|
|
# Convert to Hz
|
|
pitch = np.array(2 ** pitch)[None, :]
|
|
|
|
# Optionally return threshold
|
|
if self.return_threshold:
|
|
return pitch, np.array(threshold)
|
|
|
|
return pitch
|
|
|
|
|
|
###############################################################################
|
|
# Periodicity thresholding methods
|
|
###############################################################################
|
|
|
|
|
|
class Silence:
|
|
"""Set periodicity to zero in silent regions"""
|
|
|
|
def __init__(self, value=-60):
|
|
self.value = value
|
|
|
|
def __call__(self,
|
|
periodicity,
|
|
audio,
|
|
sample_rate=onnxcrepe.SAMPLE_RATE,
|
|
precision=None,
|
|
pad=True):
|
|
# Don't modify in-place
|
|
periodicity = periodicity.copy()
|
|
|
|
# Compute loudness
|
|
hop_length = sample_rate * precision // 1000
|
|
loudness = onnxcrepe.loudness.a_weighted(
|
|
audio, sample_rate, hop_length, pad)
|
|
|
|
# Threshold silence
|
|
periodicity[loudness < self.value] = 0.
|
|
|
|
return periodicity
|