Sing-For-Me / vc_infer_pipeline.py
Jarod Castillo
comments added
9d54d5c
raw
history blame
17.9 kB
import numpy as np, parselmouth, torch, pdb, sys, os
from time import time as ttime
import torch.nn.functional as F
import scipy.signal as signal
import pyworld, os, traceback, faiss, librosa, torchcrepe
from scipy import signal
from functools import lru_cache
now_dir = os.getcwd()
sys.path.append(now_dir)
bh, ah = signal.butter(N=5, Wn=48, btype="high", fs=16000) # Design for the audio filter
input_audio_path2wav = {}
@lru_cache
def cache_harvest_f0(input_audio_path, fs, f0max, f0min, frame_period):
audio = input_audio_path2wav[input_audio_path]
f0, t = pyworld.harvest(
audio,
fs=fs,
f0_ceil=f0max,
f0_floor=f0min,
frame_period=frame_period,
)
f0 = pyworld.stonemask(audio, f0, t, fs)
return f0
def change_rms(data1, sr1, data2, sr2, rate): # 1是输入音频,2是输出音频,rate是2的占比
# print(data1.max(),data2.max())
rms1 = librosa.feature.rms(
y=data1, frame_length=sr1 // 2 * 2, hop_length=sr1 // 2
) # 每半秒一个点
rms2 = librosa.feature.rms(y=data2, frame_length=sr2 // 2 * 2, hop_length=sr2 // 2)
rms1 = torch.from_numpy(rms1)
rms1 = F.interpolate(
rms1.unsqueeze(0), size=data2.shape[0], mode="linear"
).squeeze()
rms2 = torch.from_numpy(rms2)
rms2 = F.interpolate(
rms2.unsqueeze(0), size=data2.shape[0], mode="linear"
).squeeze()
rms2 = torch.max(rms2, torch.zeros_like(rms2) + 1e-6)
data2 *= (
torch.pow(rms1, torch.tensor(1 - rate))
* torch.pow(rms2, torch.tensor(rate - 1))
).numpy()
return data2
class VC(object):
"""
Voice Conversion system.
"""
def __init__(self, tgt_sr, config):
self.x_pad, self.x_query, self.x_center, self.x_max, self.is_half = (
config.x_pad,
config.x_query,
config.x_center,
config.x_max,
config.is_half,
)
"""
The class has several parameters that get initialized such as `x_pad`, `x_query`, and so on from the configuration object provided.
These parameters appear to be related to audio processing, specifying things like sample rate, window size, padding amounts, etc.
"""
self.sr = 16000 # Hubert input sample rate
self.window = 160 # Number of points per frame
self.t_pad = self.sr * self.x_pad # Padding time before and after each segment
self.t_pad_tgt = tgt_sr * self.x_pad
self.t_pad2 = self.t_pad * 2
self.t_query = self.sr * self.x_query # Query time before and after each query point
self.t_center = self.sr * self.x_center # Query point position
self.t_max = self.sr * self.x_max # Duration threshold for non-query time
self.device = config.device
def get_f0(
self,
input_audio_path,
x,
p_len,
f0_up_key,
f0_method,
filter_radius,
inp_f0=None,
):
"""
Extracts fundamental frequency ('F0' or pitch) from a given audio signal
Multiple methods are available, such as 'pm', 'harvest', 'crepe', 'rmvpe'
Libraries 'parselmouth', 'torchcrepe' compute pitch, and 'cache_harvest_f0' is being used to compute pitch
"""
global input_audio_path2wav
time_step = self.window / self.sr * 1000
f0_min = 50
f0_max = 1100
f0_mel_min = 1127 * np.log(1 + f0_min / 700)
f0_mel_max = 1127 * np.log(1 + f0_max / 700)
if f0_method == "pm":
f0 = (
parselmouth.Sound(x, self.sr)
.to_pitch_ac(
time_step=time_step / 1000,
voicing_threshold=0.6,
pitch_floor=f0_min,
pitch_ceiling=f0_max,
)
.selected_array["frequency"]
)
pad_size = (p_len - len(f0) + 1) // 2
if pad_size > 0 or p_len - len(f0) - pad_size > 0:
f0 = np.pad(
f0, [[pad_size, p_len - len(f0) - pad_size]], mode="constant"
)
elif f0_method == "harvest":
input_audio_path2wav[input_audio_path] = x.astype(np.double)
f0 = cache_harvest_f0(input_audio_path, self.sr, f0_max, f0_min, 10)
if filter_radius > 2:
f0 = signal.medfilt(f0, 3)
elif f0_method == "crepe":
model = "full"
# Pick a batch size that doesn't cause memory errors on your gpu
batch_size = 512
# Compute pitch using first gpu
audio = torch.tensor(np.copy(x))[None].float()
f0, pd = torchcrepe.predict(
audio,
self.sr,
self.window,
f0_min,
f0_max,
model,
batch_size=batch_size,
device=self.device,
return_periodicity=True,
)
pd = torchcrepe.filter.median(pd, 3)
f0 = torchcrepe.filter.mean(f0, 3)
f0[pd < 0.1] = 0
f0 = f0[0].cpu().numpy()
elif f0_method == "rmvpe":
if hasattr(self, "model_rmvpe") == False:
from rmvpe import RMVPE
print("loading rmvpe model")
self.model_rmvpe = RMVPE(
"rmvpe.pt", is_half=self.is_half, device=self.device
)
f0 = self.model_rmvpe.infer_from_audio(x, thred=0.03)
f0 *= pow(2, f0_up_key / 12)
# with open("test.txt","w")as f:f.write("\n".join([str(i)for i in f0.tolist()]))
tf0 = self.sr // self.window # 每秒f0点数
if inp_f0 is not None:
delta_t = np.round(
(inp_f0[:, 0].max() - inp_f0[:, 0].min()) * tf0 + 1
).astype("int16")
replace_f0 = np.interp(
list(range(delta_t)), inp_f0[:, 0] * 100, inp_f0[:, 1]
)
shape = f0[self.x_pad * tf0 : self.x_pad * tf0 + len(replace_f0)].shape[0]
f0[self.x_pad * tf0 : self.x_pad * tf0 + len(replace_f0)] = replace_f0[
:shape
]
# with open("test_opt.txt","w")as f:f.write("\n".join([str(i)for i in f0.tolist()]))
f0bak = f0.copy()
f0_mel = 1127 * np.log(1 + f0 / 700)
f0_mel[f0_mel > 0] = (f0_mel[f0_mel > 0] - f0_mel_min) * 254 / (
f0_mel_max - f0_mel_min
) + 1
f0_mel[f0_mel <= 1] = 1
f0_mel[f0_mel > 255] = 255
f0_coarse = np.rint(f0_mel).astype(np.int)
return f0_coarse, f0bak # 1-0
def vc(
self,
model,
net_g,
sid,
audio0,
pitch,
pitchf, # ???
times,
index,
big_npy,
index_rate,
version,
protect,
): # ,file_index,file_big_npy
"""
The holy grail, the main conversion function.
Takes an numpy audio signal, processes it through a model, spits out a numpy audio signal.
Modifies the pitch (or 'F0') of the audio signal, given the 'pitch' and 'pitchf' parameters.
Neural network generator (net_g) infers the voice.
'index' and 'big_npy' is used to retrieve similar audio features from a pre-computed database for better conversion quality.
"""
feats = torch.from_numpy(audio0)
if self.is_half:
feats = feats.half()
else:
feats = feats.float()
if feats.dim() == 2: # double channels
feats = feats.mean(-1)
assert feats.dim() == 1, feats.dim()
feats = feats.view(1, -1)
padding_mask = torch.BoolTensor(feats.shape).to(self.device).fill_(False)
inputs = {
"source": feats.to(self.device),
"padding_mask": padding_mask,
"output_layer": 9 if version == "v1" else 12,
}
t0 = ttime()
with torch.no_grad():
logits = model.extract_features(**inputs)
feats = model.final_proj(logits[0]) if version == "v1" else logits[0]
if protect < 0.5 and pitch != None and pitchf != None:
feats0 = feats.clone()
if (
isinstance(index, type(None)) == False
and isinstance(big_npy, type(None)) == False
and index_rate != 0
):
npy = feats[0].cpu().numpy()
if self.is_half:
npy = npy.astype("float32")
# _, I = index.search(npy, 1)
# npy = big_npy[I.squeeze()]
score, ix = index.search(npy, k=8)
weight = np.square(1 / score)
weight /= weight.sum(axis=1, keepdims=True)
npy = np.sum(big_npy[ix] * np.expand_dims(weight, axis=2), axis=1)
if self.is_half:
npy = npy.astype("float16")
feats = (
torch.from_numpy(npy).unsqueeze(0).to(self.device) * index_rate
+ (1 - index_rate) * feats
)
feats = F.interpolate(feats.permute(0, 2, 1), scale_factor=2).permute(0, 2, 1)
if protect < 0.5 and pitch != None and pitchf != None:
feats0 = F.interpolate(feats0.permute(0, 2, 1), scale_factor=2).permute(
0, 2, 1
)
t1 = ttime()
p_len = audio0.shape[0] // self.window
if feats.shape[1] < p_len:
p_len = feats.shape[1]
if pitch != None and pitchf != None:
pitch = pitch[:, :p_len]
pitchf = pitchf[:, :p_len]
if protect < 0.5 and pitch != None and pitchf != None:
pitchff = pitchf.clone()
pitchff[pitchf > 0] = 1
pitchff[pitchf < 1] = protect
pitchff = pitchff.unsqueeze(-1)
feats = feats * pitchff + feats0 * (1 - pitchff)
feats = feats.to(feats0.dtype)
p_len = torch.tensor([p_len], device=self.device).long()
with torch.no_grad():
if pitch != None and pitchf != None:
audio1 = (
(net_g.infer(feats, p_len, pitch, pitchf, sid)[0][0, 0])
.data.cpu()
.float()
.numpy()
)
else:
audio1 = (
(net_g.infer(feats, p_len, sid)[0][0, 0]).data.cpu().float().numpy()
)
del feats, p_len, padding_mask
if torch.cuda.is_available():
torch.cuda.empty_cache()
t2 = ttime()
times[0] += t1 - t0
times[2] += t2 - t1
return audio1
def pipeline(
self,
model,
net_g,
sid,
audio,
input_audio_path,
times,
f0_up_key,
f0_method,
file_index,
# file_big_npy,
index_rate,
if_f0,
filter_radius,
tgt_sr,
resample_sr,
rms_mix_rate,
version,
protect,
f0_file=None,
):
"""
This is a pipeline function that strings together multiple operations for voice conversion.
The function does some preprocessing on the input audio(e.g. filtering)
The function then segments the audio into pieces and processes each segment through the voice conversion ('vc') method
The converted segments are then concatenated to produce the final converted audio
"""
# Phase 1: Load index file
if (
file_index != ""
# and file_big_npy != ""
# and os.path.exists(file_big_npy) == True
and os.path.exists(file_index) == True
and index_rate != 0
):
try:
index = faiss.read_index(file_index) # Read from the vector store
# big_npy = np.load(file_big_npy)
big_npy = index.reconstruct_n(0, index.ntotal) # Convert index into a big numpy array
except:
traceback.print_exc()
index = big_npy = None
else:
index = big_npy = None # If we don't have the index file, it's ok we won't use it.
# Phase 2: Filter audio signal
audio = signal.filtfilt(bh, ah, audio) # Avoid phase distortion
audio_pad = np.pad(audio, (self.window // 2, self.window // 2), mode="reflect") # Padding to ensure we calculate the beginning and end accurately
optimal_time_shifts = []
if audio_pad.shape[0] > self.t_max:
audio_sum = np.zeros_like(audio) # Create numpy array filled with zeros and same shape as audio
for i in range(self.window):
audio_sum += audio_pad[i : i - self.window]
for t in range(self.t_center, audio.shape[0], self.t_center):
optimal_time_shifts.append(
t
- self.t_query
+ np.where(
np.abs(audio_sum[t - self.t_query : t + self.t_query])
== np.abs(audio_sum[t - self.t_query : t + self.t_query]).min()
)[0][0]
)
s = 0
audio_opt = []
t = None
t1 = ttime()
audio_pad = np.pad(audio, (self.t_pad, self.t_pad), mode="reflect") # Reflection of the signal's edges on both ends
number_of_analysis_frames = audio_pad.shape[0] // self.window
inp_f0 = None
if hasattr(f0_file, "name") == True:
try:
with open(f0_file.name, "r") as f:
lines = f.read().strip("\n").split("\n")
inp_f0 = []
for line in lines:
inp_f0.append([float(i) for i in line.split(",")])
inp_f0 = np.array(inp_f0, dtype="float32")
except:
traceback.print_exc()
sid = torch.tensor(sid, device=self.device).unsqueeze(0).long()
pitch, pitchf = None, None
if if_f0 == 1:
# Calls the `get_f0` method to calculate pitch values based on audio features.
# These calculated pitch values are used for voice conversion.
pitch, pitchf = self.get_f0(
input_audio_path,
audio_pad,
number_of_analysis_frames,
f0_up_key,
f0_method,
filter_radius,
inp_f0,
)
pitch = pitch[:number_of_analysis_frames]
pitchf = pitchf[:number_of_analysis_frames]
if self.device == "mps":
pitchf = pitchf.astype(np.float32)
pitch = torch.tensor(pitch, device=self.device).unsqueeze(0).long()
pitchf = torch.tensor(pitchf, device=self.device).unsqueeze(0).float()
t2 = ttime()
times[1] += t2 - t1
for t in optimal_time_shifts:
t = t // self.window * self.window
if if_f0 == 1:
audio_opt.append(
self.vc(
model,
net_g,
sid,
audio_pad[s : t + self.t_pad2 + self.window],
pitch[:, s // self.window : (t + self.t_pad2) // self.window],
pitchf[:, s // self.window : (t + self.t_pad2) // self.window],
times,
index,
big_npy,
index_rate,
version,
protect,
)[self.t_pad_tgt : -self.t_pad_tgt]
)
else:
audio_opt.append(
self.vc(
model,
net_g,
sid,
audio_pad[s : t + self.t_pad2 + self.window],
None,
None,
times,
index,
big_npy,
index_rate,
version,
protect,
)[self.t_pad_tgt : -self.t_pad_tgt]
)
s = t
if if_f0 == 1:
audio_opt.append(
self.vc(
model,
net_g,
sid,
audio_pad[t:],
pitch[:, t // self.window :] if t is not None else pitch,
pitchf[:, t // self.window :] if t is not None else pitchf,
times,
index,
big_npy,
index_rate,
version,
protect,
)[self.t_pad_tgt : -self.t_pad_tgt]
)
else:
audio_opt.append(
self.vc(
model,
net_g,
sid,
audio_pad[t:],
None,
None,
times,
index,
big_npy,
index_rate,
version,
protect,
)[self.t_pad_tgt : -self.t_pad_tgt]
)
audio_opt = np.concatenate(audio_opt)
if rms_mix_rate != 1:
audio_opt = change_rms(audio, 16000, audio_opt, tgt_sr, rms_mix_rate)
if resample_sr >= 16000 and tgt_sr != resample_sr:
audio_opt = librosa.resample(
audio_opt, orig_sr=tgt_sr, target_sr=resample_sr
)
audio_max = np.abs(audio_opt).max() / 0.99
max_int16 = 32768
if audio_max > 1:
max_int16 /= audio_max
audio_opt = (audio_opt * max_int16).astype(np.int16)
del pitch, pitchf, sid
if torch.cuda.is_available():
torch.cuda.empty_cache()
return audio_opt