File size: 9,806 Bytes
28d6785
 
 
 
 
 
 
 
 
 
 
 
546e179
 
28d6785
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
087c3c2
 
28d6785
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
087c3c2
28d6785
 
 
 
087c3c2
28d6785
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
087c3c2
28d6785
087c3c2
 
 
 
 
 
 
28d6785
 
 
 
 
 
 
 
 
 
 
 
 
 
 
087c3c2
 
 
 
 
 
 
28d6785
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3971395
28d6785
 
3ec7c97
28d6785
 
 
 
 
3971395
28d6785
 
 
 
5d57342
3971395
28d6785
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
087c3c2
28d6785
 
 
 
d6287b5
c31d772
28d6785
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5d57342
6bdadbb
 
 
 
 
 
 
 
 
 
28d6785
 
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
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
import asyncio
import datetime
import logging
import os
import time
import traceback

import edge_tts
import gradio as gr
import librosa
import torch
from fairseq import checkpoint_utils
from huggingface_hub import snapshot_download


from config import Config
from lib.infer_pack.models import (
    SynthesizerTrnMs256NSFsid,
    SynthesizerTrnMs256NSFsid_nono,
    SynthesizerTrnMs768NSFsid,
    SynthesizerTrnMs768NSFsid_nono,
)
from rmvpe import RMVPE
from vc_infer_pipeline import VC

logging.getLogger("fairseq").setLevel(logging.WARNING)
logging.getLogger("numba").setLevel(logging.WARNING)
logging.getLogger("markdown_it").setLevel(logging.WARNING)
logging.getLogger("urllib3").setLevel(logging.WARNING)
logging.getLogger("matplotlib").setLevel(logging.WARNING)

limitation = os.getenv("SYSTEM") == "spaces"

config = Config()

# Edge TTS
edge_output_filename = "edge_output.mp3"
tts_voice_list = asyncio.get_event_loop().run_until_complete(edge_tts.list_voices())
tts_voices = [f"{v['ShortName']}-{v['Gender']}" for v in tts_voice_list]

# RVC models
model_root = snapshot_download(repo_id="NoCrypt/miku_RVC", token=os.environ["TOKEN"])
models = [d for d in os.listdir(model_root) if os.path.isdir(f"{model_root}/{d}")]
models.sort()


def model_data(model_name):
    # global n_spk, tgt_sr, net_g, vc, cpt, version, index_file
    pth_path = [
        f"{model_root}/{model_name}/{f}"
        for f in os.listdir(f"{model_root}/{model_name}")
        if f.endswith(".pth")
    ][0]
    print(f"Loading {pth_path}")
    cpt = torch.load(pth_path, map_location="cpu")
    tgt_sr = cpt["config"][-1]
    cpt["config"][-3] = cpt["weight"]["emb_g.weight"].shape[0]  # n_spk
    if_f0 = cpt.get("f0", 1)
    version = cpt.get("version", "v1")
    if version == "v1":
        if if_f0 == 1:
            net_g = SynthesizerTrnMs256NSFsid(*cpt["config"], is_half=config.is_half)
        else:
            net_g = SynthesizerTrnMs256NSFsid_nono(*cpt["config"])
    elif version == "v2":
        if if_f0 == 1:
            net_g = SynthesizerTrnMs768NSFsid(*cpt["config"], is_half=config.is_half)
        else:
            net_g = SynthesizerTrnMs768NSFsid_nono(*cpt["config"])
    else:
        raise ValueError("Unknown version")
    del net_g.enc_q
    net_g.load_state_dict(cpt["weight"], strict=False)
    print("Model loaded")
    net_g.eval().to(config.device)
    if config.is_half:
        net_g = net_g.half()
    else:
        net_g = net_g.float()
    vc = VC(tgt_sr, config)
    # n_spk = cpt["config"][-3]

    index_files = [
        f"{model_root}/{model_name}/{f}"
        for f in os.listdir(f"{model_root}/{model_name}")
        if f.endswith(".index")
    ]
    if len(index_files) == 0:
        print("No index file found")
        index_file = ""
    else:
        index_file = index_files[0]
        print(f"Index file found: {index_file}")

    return tgt_sr, net_g, vc, version, index_file, if_f0


def load_hubert():
    # global hubert_model
    models, _, _ = checkpoint_utils.load_model_ensemble_and_task(
        ["hubert_base.pt"],
        suffix="",
    )
    hubert_model = models[0]
    hubert_model = hubert_model.to(config.device)
    if config.is_half:
        hubert_model = hubert_model.half()
    else:
        hubert_model = hubert_model.float()
    return hubert_model.eval()


def tts(
    model_name,
    speed,
    tts_text,
    tts_voice,
    f0_up_key,
    f0_method,
    index_rate,
    protect,
    filter_radius=3,
    resample_sr=0,
    rms_mix_rate=0.25,
):
    print("------------------")
    print(datetime.datetime.now())
    print("tts_text:")
    print(tts_text)
    print(f"tts_voice: {tts_voice}, speed: {speed}")
    print(f"Model name: {model_name}")
    print(f"F0: {f0_method}, Key: {f0_up_key}, Index: {index_rate}, Protect: {protect}")
    try:
        if limitation and len(tts_text) > 1000:
            print("Error: Text too long")
            return (
                f"Text characters should be at most 1000 in this huggingface space, but got {len(tts_text)} characters.",
                None,
                None,
            )
        t0 = time.time()
        if speed >= 0:
            speed_str = f"+{speed}%"
        else:
            speed_str = f"{speed}%"
        asyncio.run(
            edge_tts.Communicate(
                tts_text, "-".join(tts_voice.split("-")[:-1]), rate=speed_str
            ).save(edge_output_filename)
        )
        t1 = time.time()
        edge_time = t1 - t0
        audio, sr = librosa.load(edge_output_filename, sr=16000, mono=True)
        duration = len(audio) / sr
        print(f"Audio duration: {duration}s")
        if limitation and duration >= 200:
            print("Error: Audio too long")
            return (
                f"Audio should be less than 200 seconds in this huggingface space, but got {duration}s.",
                edge_output_filename,
                None,
            )
        f0_up_key = int(f0_up_key)

        tgt_sr, net_g, vc, version, index_file, if_f0 = model_data(model_name)
        if f0_method == "rmvpe":
            vc.model_rmvpe = rmvpe_model
        times = [0, 0, 0]
        audio_opt = vc.pipeline(
            hubert_model,
            net_g,
            0,
            audio,
            edge_output_filename,
            times,
            f0_up_key,
            f0_method,
            index_file,
            # file_big_npy,
            index_rate,
            if_f0,
            filter_radius,
            tgt_sr,
            resample_sr,
            rms_mix_rate,
            version,
            protect,
            None,
        )
        if tgt_sr != resample_sr >= 16000:
            tgt_sr = resample_sr
        info = f"Success. Time: edge-tts: {edge_time}s, npy: {times[0]}s, f0: {times[1]}s, infer: {times[2]}s"
        print(info)
        return (
            info,
            edge_output_filename,
            (tgt_sr, audio_opt),
        )
    except EOFError:
        info = (
            "It seems that the edge-tts output is not valid. "
            "This may occur when the input text and the speaker do not match. "
            "For example, maybe you entered Japanese (without alphabets) text but chose non-Japanese speaker?"
        )
        print(info)
        return info, None, None
    except:
        info = traceback.format_exc()
        print(info)
        return info, None, None


print("Loading hubert model...")
hubert_model = load_hubert()
print("Hubert model loaded.")

print("Loading rmvpe model...")
rmvpe_model = RMVPE("rmvpe.pt", config.is_half, config.device)
print("rmvpe model loaded.")

initial_md = """
![banner that says mikutts](https://huggingface.co/spaces/NoCrypt/mikuTTS/resolve/main/imgs/banner_mikutts.webp)
"""

app = gr.Blocks(theme='NoCrypt/miku')
with app:
    gr.Markdown(initial_md)
    with gr.Row():
        with gr.Column():
            model_name = gr.Dropdown(
                label="Model",
                choices=models,
                value=models[0],
            )
            f0_key_up = gr.Number(
                label="Tune",
                value=6,
            )
        with gr.Column():
            f0_method = gr.Radio(
                label="Pitch extraction method (pm: very fast, low quality, rmvpe: a little slow, high quality)",
                choices=["pm", "rmvpe"],  # harvest and crepe is too slow
                value="rmvpe",
                interactive=True,
            )
            index_rate = gr.Slider(
                minimum=0,
                maximum=1,
                label="Index rate",
                value=1,
                interactive=True,
            )
            protect0 = gr.Slider(
                minimum=0,
                maximum=0.5,
                label="Protect",
                value=0.33,
                step=0.01,
                interactive=True,
            )
    with gr.Row():
        with gr.Column():
            tts_voice = gr.Dropdown(
                label="Edge-tts speaker (format: language-Country-Name-Gender), make sure the gender matches the model",
                choices=tts_voices,
                allow_custom_value=False,
                value="ja-JP-NanamiNeural-Female",
            )
            speed = gr.Slider(
                minimum=-100,
                maximum=100,
                label="Speech speed (%)",
                value=0,
                step=10,
                interactive=True,
            )
            tts_text = gr.Textbox(label="Input Text", value="こんにちは、私の名前は初音ミクです!")
        with gr.Column():
            but0 = gr.Button("Convert", variant="primary")
            info_text = gr.Textbox(label="Output info")
        with gr.Column():
            with gr.Accordion("Edge Voice", open=False):
                edge_tts_output = gr.Audio(label="Edge Voice", type="filepath")
            tts_output = gr.Audio(label="Result")
        but0.click(
            tts,
            [
                model_name,
                speed,
                tts_text,
                tts_voice,
                f0_key_up,
                f0_method,
                index_rate,
                protect0,
            ],
            [info_text, edge_tts_output, tts_output],
        )
    with gr.Row():
        examples = gr.Examples(
            examples_per_page=100,
            examples=[
                ["こんにちは、私の名前は初音ミクです!", "ja-JP-NanamiNeural-Female", 6],
                ["Hello there. My name is Hatsune Miku!","en-CA-ClaraNeural-Female", 6],
                ["Halo. Nama saya Hatsune Miku!","id-ID-GadisNeural-Female", 4],
                ["Halo. Jenengku Hatsune Miku!","jv-ID-SitiNeural-Female", 10],
            ],
            inputs=[tts_text, tts_voice, f0_key_up],
        )

app.launch()