Hugging Face's logo

Spaces:
LosCaquitos
/
RVC-CH
Running

App Files Community
RVC-CH
File size: 7,171 Bytes
8be9381
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
754f043
8be9381
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
import logging
import os
import pathlib
import re
import sys
import unicodedata
import warnings

import soxr

import wget

import numpy as np

from torch import nn
from transformers import HubertModel

import librosa
import soundfile as sf

from lib.rvc.common import RVC_MODELS_DIR

# Remove this to see warnings about transformers models
warnings.filterwarnings("ignore")

logging.getLogger("fairseq").setLevel(logging.ERROR)
logging.getLogger("faiss.loader").setLevel(logging.ERROR)
logging.getLogger("transformers").setLevel(logging.ERROR)
logging.getLogger("torch").setLevel(logging.ERROR)

now_dir = pathlib.Path.cwd()
sys.path.append(str(now_dir))

base_path = os.path.join(str(RVC_MODELS_DIR), "formant", "stftpitchshift")
stft = base_path + ".exe" if sys.platform == "win32" else base_path


class HubertModelWithFinalProj(HubertModel):
    def __init__(self, config):
        super().__init__(config)
        self.final_proj = nn.Linear(config.hidden_size, config.classifier_proj_size)


def load_audio_16k(file):
    # this is used by f0 and feature extractions that load preprocessed 16k files, so there's no need to resample
    try:
        audio, sr = librosa.load(file, sr=16000)
    except Exception as error:
        raise RuntimeError(f"An error occurred loading the audio: {error}")

    return audio.flatten()


def load_audio(file, sample_rate):
    try:
        file = file.strip(" ").strip('"').strip("\n").strip('"').strip(" ")
        audio, sr = sf.read(file)
        if len(audio.shape) > 1:
            audio = librosa.to_mono(audio.T)
        if sr != sample_rate:
            audio = librosa.resample(
                audio,
                orig_sr=sr,
                target_sr=sample_rate,
                res_type="soxr_vhq",
            )
    except Exception as error:
        raise RuntimeError(f"An error occurred loading the audio: {error}")

    return audio.flatten()


def load_audio_infer(
    file,
    sample_rate,
    **kwargs,
):
    formant_shifting = kwargs.get("formant_shifting", False)
    try:
        file = file.strip(" ").strip('"').strip("\n").strip('"').strip(" ")
        if not pathlib.Path(file).is_file():
            raise FileNotFoundError(f"File not found: {file}")
        audio, sr = sf.read(file)
        if len(audio.shape) > 1:
            audio = librosa.to_mono(audio.T)
        if sr != sample_rate:
            audio = librosa.resample(
                audio,
                orig_sr=sr,
                target_sr=sample_rate,
                res_type="soxr_vhq",
            )
        if formant_shifting:
            formant_qfrency = kwargs.get("formant_qfrency", 0.8)
            formant_timbre = kwargs.get("formant_timbre", 0.8)

            from stftpitchshift import StftPitchShift

            pitchshifter = StftPitchShift(1024, 32, sample_rate)
            audio = pitchshifter.shiftpitch(
                audio,
                factors=1,
                quefrency=formant_qfrency * 1e-3,
                distortion=formant_timbre,
            )
    except Exception as error:
        raise RuntimeError(f"An error occurred loading the audio: {error}")
    return np.array(audio).flatten()


def format_title(title):
    formatted_title = unicodedata.normalize("NFC", title)
    formatted_title = re.sub(r"[\u2500-\u257F]+", "", formatted_title)
    formatted_title = re.sub(r"[^\w\s.-]", "", formatted_title, flags=re.UNICODE)
    formatted_title = re.sub(r"\s+", "_", formatted_title)
    return formatted_title


def load_embedding(embedder_model, custom_embedder=None):
    embedder_root = os.path.join(str(RVC_MODELS_DIR), "embedders")
    embedding_list = {
        "contentvec": os.path.join(embedder_root, "contentvec"),
        "spin": os.path.join(embedder_root, "spin"),
        "spin-v2": os.path.join(embedder_root, "spin-v2"),
        "chinese-hubert-base": os.path.join(embedder_root, "chinese_hubert_base"),
        "japanese-hubert-base": os.path.join(embedder_root, "japanese_hubert_base"),
        "korean-hubert-base": os.path.join(embedder_root, "korean_hubert_base"),
    }

    online_embedders = {
        "contentvec": (
            "https://huggingface.co/JackismyShephard/ultimate-rvc/resolve/main/Resources/embedders/contentvec/pytorch_model.bin"
        ),
        "spin": (
            "https://huggingface.co/JackismyShephard/ultimate-rvc/resolve/main/Resources/embedders/spin/pytorch_model.bin"
        ),
        "spin-v2": (
            "https://huggingface.co/JackismyShephard/ultimate-rvc/resolve/main/Resources/embedders/spin-v2/pytorch_model.bin"
        ),
        "chinese-hubert-base": (
            "https://huggingface.co/JackismyShephard/ultimate-rvc/resolve/main/Resources/embedders/chinese_hubert_base/pytorch_model.bin"
        ),
        "japanese-hubert-base": (
            "https://huggingface.co/JackismyShephard/ultimate-rvc/resolve/main/Resources/embedders/japanese_hubert_base/pytorch_model.bin"
        ),
        "korean-hubert-base": (
            "https://huggingface.co/JackismyShephard/ultimate-rvc/resolve/main/Resources/embedders/korean_hubert_base/pytorch_model.bin"
        ),
    }

    config_files = {
        "contentvec": (
            "https://huggingface.co/JackismyShephard/ultimate-rvc/resolve/main/Resources/embedders/contentvec/config.json"
        ),
        "spin": (
            "https://huggingface.co/JackismyShephard/ultimate-rvc/resolve/main/Resources/embedders/spin/config.json"
        ),
        "spin-v2": (
            "https://huggingface.co/JackismyShephard/ultimate-rvc/resolve/main/Resources/embedders/spin-v2/config.json"
        ),
        "chinese-hubert-base": (
            "https://huggingface.co/JackismyShephard/ultimate-rvc/resolve/main/Resources/embedders/chinese_hubert_base/config.json"
        ),
        "japanese-hubert-base": (
            "https://huggingface.co/JackismyShephard/ultimate-rvc/resolve/main/Resources/embedders/japanese_hubert_base/config.json"
        ),
        "korean-hubert-base": (
            "https://huggingface.co/JackismyShephard/ultimate-rvc/resolve/main/Resources/embedders/korean_hubert_base/config.json"
        ),
    }

    if embedder_model == "custom":
        if pathlib.Path(custom_embedder).exists():
            model_path = custom_embedder
        else:
            print(f"Custom embedder not found: {custom_embedder}, using contentvec")
            model_path = embedding_list["contentvec"]
    else:
        model_path = embedding_list[embedder_model]
        bin_file = os.path.join(model_path, "pytorch_model.bin")
        json_file = os.path.join(model_path, "config.json")
        pathlib.Path(model_path).mkdir(exist_ok=True, parents=True)
        if not pathlib.Path(bin_file).exists():
            url = online_embedders[embedder_model]
            print(f"Downloading {url} to {model_path}...")
            wget.download(url, out=bin_file)
        if not pathlib.Path(json_file).exists():
            url = config_files[embedder_model]
            print(f"Downloading {url} to {model_path}...")
            wget.download(url, out=json_file)

    models = HubertModelWithFinalProj.from_pretrained(model_path)
    return models