ailia-models/diffusion/latentsync/audio2feature.py at master · ailia-ai/ailia-models · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
# Adapted from https://github.com/TMElyralab/MuseTalk/blob/main/musetalk/whisper/audio2feature.py

import numpy as np

from whisper.transcribe import transcribe as transcribe_function


class Audio2Feature:
    def __init__(
        self,
        model,
        num_frames=16,
        use_onnx: bool = False,
    ):
        self.model = model
        self.num_frames = num_frames
        self.embedding_dim = 384
        self.use_onnx = use_onnx

    def get_sliced_feature(
        self, feature_array, vid_idx, audio_feat_length=[2, 2], fps=25
    ):
        """
        Get sliced features based on a given index
        :param feature_array:
        :param start_idx: the start index of the feature
        :param audio_feat_length:
        :return:
        """
        length = len(feature_array)
        selected_feature = []
        selected_idx = []

        center_idx = int(vid_idx * 50 / fps)
        left_idx = center_idx - audio_feat_length[0] * 2
        right_idx = center_idx + (audio_feat_length[1] + 1) * 2

        for idx in range(left_idx, right_idx):
            idx = max(0, idx)
            idx = min(length - 1, idx)
            x = feature_array[idx]
            selected_feature.append(x)
            selected_idx.append(idx)

        selected_feature = np.concatenate(selected_feature, axis=0)
        selected_feature = selected_feature.reshape(-1, self.embedding_dim)  # 50*384
        return selected_feature, selected_idx

    def feature2chunks(self, feature_array, fps, audio_feat_length=[2, 2]):
        whisper_chunks = []
        whisper_idx_multiplier = 50.0 / fps

        i = 0
        while True:
            start_idx = int(i * whisper_idx_multiplier)
            selected_feature, _ = self.get_sliced_feature(
                feature_array=feature_array,
                vid_idx=i,
                audio_feat_length=audio_feat_length,
                fps=fps,
            )
            whisper_chunks.append(selected_feature)
            i += 1
            if start_idx > len(feature_array):
                break

        return whisper_chunks

    def audio2feat(self, audio):
        # get the sample rate of the audio
        result = transcribe_function(self.model, audio, use_onnx=self.use_onnx)
        embed_list = []
        for emb in result["segments"]:
            encoder_embeddings = emb["encoder_embeddings"]
            encoder_embeddings = encoder_embeddings.transpose(0, 2, 1, 3)
            encoder_embeddings = encoder_embeddings.squeeze(0)
            start_idx = int(emb["start"])
            end_idx = int(emb["end"])
            emb_end_idx = int((end_idx - start_idx) / 2)
            embed_list.append(encoder_embeddings[:emb_end_idx])

        concatenated_array = np.concatenate(embed_list, axis=0)
        return concatenated_array