Upload 5 files

difpoint/dataset_process/audio.py

@@ -0,0 +1,156 @@
+import librosa
+import librosa.filters
+import numpy as np
+# import tensorflow as tf
+from scipy import signal
+from scipy.io import wavfile
+from difpoint.src.utils.hparams import hparams as hp
+
+
+def load_wav(path, sr):
+    return librosa.core.load(path, sr=sr)[0]
+
+
+def save_wav(wav, path, sr):
+    wav *= 32767 / max(0.01, np.max(np.abs(wav)))
+    # proposed by @dsmiller
+    wavfile.write(path, sr, wav.astype(np.int16))
+
+
+def save_wavenet_wav(wav, path, sr):
+    librosa.output.write_wav(path, wav, sr=sr)
+
+
+def preemphasis(wav, k, preemphasize=True):
+    if preemphasize:
+        return signal.lfilter([1, -k], [1], wav)
+    return wav
+
+
+def inv_preemphasis(wav, k, inv_preemphasize=True):
+    if inv_preemphasize:
+        return signal.lfilter([1], [1, -k], wav)
+    return wav
+
+
+def get_hop_size():
+    hop_size = hp.hop_size
+    if hop_size is None:
+        assert hp.frame_shift_ms is not None
+        hop_size = int(hp.frame_shift_ms / 1000 * hp.sample_rate)
+    return hop_size
+
+
+def linearspectrogram(wav):
+    D = _stft(preemphasis(wav, hp.preemphasis, hp.preemphasize))
+    S = _amp_to_db(np.abs(D)) - hp.ref_level_db
+
+    if hp.signal_normalization:
+        return _normalize(S)
+    return S
+
+
+def melspectrogram(wav):
+    D = _stft(preemphasis(wav, hp.preemphasis, hp.preemphasize))
+    S = _amp_to_db(_linear_to_mel(np.abs(D))) - hp.ref_level_db
+
+    if hp.signal_normalization:
+        return _normalize(S)
+    return S
+
+
+def _lws_processor():
+    import lws
+    return lws.lws(hp.n_fft, get_hop_size(), fftsize=hp.win_size, mode="speech")
+
+
+def _stft(y):
+    if hp.use_lws:
+        return _lws_processor(hp).stft(y).T
+    else:
+        return librosa.stft(y=y, n_fft=hp.n_fft, hop_length=get_hop_size(), win_length=hp.win_size)
+
+
+##########################################################
+# Those are only correct when using lws!!! (This was messing with Wavenet quality for a long time!)
+def num_frames(length, fsize, fshift):
+    """Compute number of time frames of spectrogram
+    """
+    pad = (fsize - fshift)
+    if length % fshift == 0:
+        M = (length + pad * 2 - fsize) // fshift + 1
+    else:
+        M = (length + pad * 2 - fsize) // fshift + 2
+    return M
+
+
+def pad_lr(x, fsize, fshift):
+    """Compute left and right padding
+    """
+    M = num_frames(len(x), fsize, fshift)
+    pad = (fsize - fshift)
+    T = len(x) + 2 * pad
+    r = (M - 1) * fshift + fsize - T
+    return pad, pad + r
+
+
+##########################################################
+# Librosa correct padding
+def librosa_pad_lr(x, fsize, fshift):
+    return 0, (x.shape[0] // fshift + 1) * fshift - x.shape[0]
+
+
+# Conversions
+_mel_basis = None
+
+
+def _linear_to_mel(spectogram):
+    global _mel_basis
+    if _mel_basis is None:
+        _mel_basis = _build_mel_basis()
+    return np.dot(_mel_basis, spectogram)
+
+
+def _build_mel_basis():
+    assert hp.fmax <= hp.sample_rate // 2
+    return librosa.filters.mel(sr=hp.sample_rate, n_fft=hp.n_fft, n_mels=hp.num_mels,
+                               fmin=hp.fmin, fmax=hp.fmax)
+
+
+def _amp_to_db(x):
+    min_level = np.exp(hp.min_level_db / 20 * np.log(10))
+    return 20 * np.log10(np.maximum(min_level, x))
+
+
+def _db_to_amp(x):
+    return np.power(10.0, (x) * 0.05)
+
+
+def _normalize(S):
+    if hp.allow_clipping_in_normalization:
+        if hp.symmetric_mels:
+            return np.clip((2 * hp.max_abs_value) * ((S - hp.min_level_db) / (-hp.min_level_db)) - hp.max_abs_value,
+                           -hp.max_abs_value, hp.max_abs_value)
+        else:
+            return np.clip(hp.max_abs_value * ((S - hp.min_level_db) / (-hp.min_level_db)), 0, hp.max_abs_value)
+
+    assert S.max() <= 0 and S.min() - hp.min_level_db >= 0
+    if hp.symmetric_mels:
+        return (2 * hp.max_abs_value) * ((S - hp.min_level_db) / (-hp.min_level_db)) - hp.max_abs_value
+    else:
+        return hp.max_abs_value * ((S - hp.min_level_db) / (-hp.min_level_db))
+
+
+def _denormalize(D):
+    if hp.allow_clipping_in_normalization:
+        if hp.symmetric_mels:
+            return (((np.clip(D, -hp.max_abs_value,
+                              hp.max_abs_value) + hp.max_abs_value) * -hp.min_level_db / (2 * hp.max_abs_value))
+                    + hp.min_level_db)
+        else:
+            return ((np.clip(D, 0, hp.max_abs_value) * -hp.min_level_db / hp.max_abs_value) + hp.min_level_db)
+
+    if hp.symmetric_mels:
+        return (((D + hp.max_abs_value) * -hp.min_level_db / (2 * hp.max_abs_value)) + hp.min_level_db)
+    else:
+        return ((D * -hp.min_level_db / hp.max_abs_value) + hp.min_level_db)

difpoint/dataset_process/wav2lip.py

@@ -0,0 +1,75 @@
+import torch
+import torch.nn.functional as F
+from torch import nn
+
+
+class Conv2d(nn.Module):
+    def __init__(self, cin, cout, kernel_size, stride, padding, residual=False, use_act=True, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self.conv_block = nn.Sequential(
+            nn.Conv2d(cin, cout, kernel_size, stride, padding),
+            nn.BatchNorm2d(cout)
+        )
+        self.act = nn.ReLU()
+        self.residual = residual
+        self.use_act = use_act
+
+    def forward(self, x):
+        out = self.conv_block(x)
+        if self.residual:
+            out += x
+
+        if self.use_act:
+            return self.act(out)
+        else:
+            return out
+
+class AudioEncoder(nn.Module):
+    def __init__(self, wav2lip_checkpoint, device):
+        super(AudioEncoder, self).__init__()
+
+        self.audio_encoder = nn.Sequential(
+            Conv2d(1, 32, kernel_size=3, stride=1, padding=1),
+            Conv2d(32, 32, kernel_size=3, stride=1, padding=1, residual=True),
+            Conv2d(32, 32, kernel_size=3, stride=1, padding=1, residual=True),
+
+            Conv2d(32, 64, kernel_size=3, stride=(3, 1), padding=1),
+            Conv2d(64, 64, kernel_size=3, stride=1, padding=1, residual=True),
+            Conv2d(64, 64, kernel_size=3, stride=1, padding=1, residual=True),
+
+            Conv2d(64, 128, kernel_size=3, stride=3, padding=1),
+            Conv2d(128, 128, kernel_size=3, stride=1, padding=1, residual=True),
+            Conv2d(128, 128, kernel_size=3, stride=1, padding=1, residual=True),
+
+            Conv2d(128, 256, kernel_size=3, stride=(3, 2), padding=1),
+            Conv2d(256, 256, kernel_size=3, stride=1, padding=1, residual=True),
+
+            Conv2d(256, 512, kernel_size=3, stride=1, padding=0),
+            Conv2d(512, 512, kernel_size=1, stride=1, padding=0),)
+
+        #### load the pre-trained audio_encoder
+        wav2lip_state_dict = torch.load(wav2lip_checkpoint, map_location=torch.device(device))['state_dict']
+        state_dict = self.audio_encoder.state_dict()
+
+        for k,v in wav2lip_state_dict.items():
+            if 'audio_encoder' in k:
+                state_dict[k.replace('module.audio_encoder.', '')] = v
+        self.audio_encoder.load_state_dict(state_dict)
+
+
+    def forward(self, audio_sequences):
+        # audio_sequences = (B, T, 1, 80, 16)
+        B = audio_sequences.size(0)
+
+        audio_sequences = torch.cat([audio_sequences[:, i] for i in range(audio_sequences.size(1))], dim=0)
+
+        audio_embedding = self.audio_encoder(audio_sequences) # B, 512, 1, 1
+        dim = audio_embedding.shape[1]
+        audio_embedding = audio_embedding.reshape((B, -1, dim, 1, 1))
+
+        return audio_embedding.squeeze(-1).squeeze(-1) #B seq_len+1 512
+
+wav2lip_checkpoint='ckpts/wav2lip.pth'
+wav2lip_model = AudioEncoder(wav2lip_checkpoint, 'cuda')
+wav2lip_model.cuda()
+wav2lip_model.eval()