File size: 6,728 Bytes
cf0491a
9429d2d
cefd4f0
cf0491a
 
 
 
 
 
 
 
 
 
 
9429d2d
 
 
 
 
 
 
 
 
 
 
cf0491a
9429d2d
 
cf0491a
 
9429d2d
 
 
 
 
 
 
 
 
 
 
 
 
 
cf0491a
 
9429d2d
 
 
 
 
 
 
 
 
 
 
cf0491a
 
9429d2d
 
 
 
 
 
 
 
cf0491a
9429d2d
cf0491a
9429d2d
 
cf0491a
9429d2d
 
 
 
 
 
 
 
 
 
 
 
 
cf0491a
db71199
 
9429d2d
cf0491a
 
 
 
 
 
 
 
 
 
 
 
 
cefd4f0
9429d2d
 
 
 
 
 
 
 
 
 
db71199
cf0491a
 
 
 
 
 
 
9429d2d
cf0491a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import sys
import logging
import os
import json
import torch
import argparse
import commons
import utils
import gradio as gr

from models import SynthesizerTrn
from text.symbols import symbols
from text import cleaned_text_to_sequence, get_bert
from text.cleaner import clean_text

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)

logging.basicConfig(
    level=logging.INFO, format="| %(name)s | %(levelname)s | %(message)s"
)

logger = logging.getLogger(__name__)
limitation = os.getenv("SYSTEM") == "spaces"  # limit text and audio length in huggingface spaces


def get_text(text, hps):
    language_str = "JP"
    norm_text, phone, tone, word2ph = clean_text(text, language_str)
    phone, tone, language = cleaned_text_to_sequence(phone, tone, language_str)

    if hps.data.add_blank:
        phone = commons.intersperse(phone, 0)
        tone = commons.intersperse(tone, 0)
        language = commons.intersperse(language, 0)
        for i in range(len(word2ph)):
            word2ph[i] = word2ph[i] * 2
        word2ph[0] += 1
    bert = get_bert(norm_text, word2ph, language_str, device)
    del word2ph
    assert bert.shape[-1] == len(phone), phone

    ja_bert = bert
    bert = torch.zeros(1024, len(phone))

    assert bert.shape[-1] == len(
        phone
    ), f"Bert seq len {bert.shape[-1]} != {len(phone)}"

    phone = torch.LongTensor(phone)
    tone = torch.LongTensor(tone)
    language = torch.LongTensor(language)
    return bert, ja_bert, phone, tone, language


def infer(text, sdp_ratio, noise_scale, noise_scale_w, length_scale, sid, net_g_ms, hps):
    bert, ja_bert, phones, tones, lang_ids = get_text(text, hps)
    with torch.no_grad():
        x_tst = phones.to(device).unsqueeze(0)
        tones = tones.to(device).unsqueeze(0)
        lang_ids = lang_ids.to(device).unsqueeze(0)
        bert = bert.to(device).unsqueeze(0)
        ja_bert = ja_bert.to(device).unsqueeze(0)
        x_tst_lengths = torch.LongTensor([phones.size(0)]).to(device)
        del phones
        sid = torch.LongTensor([sid]).to(device)
        audio = (
            net_g_ms.infer(
                x_tst,
                x_tst_lengths,
                sid,
                tones,
                lang_ids,
                bert,
                ja_bert,
                sdp_ratio=sdp_ratio,
                noise_scale=noise_scale,
                noise_scale_w=noise_scale_w,
                length_scale=length_scale,
            )[0][0, 0]
            .data.cpu()
            .float()
            .numpy()
        )
        del x_tst, tones, lang_ids, bert, x_tst_lengths, sid
        torch.cuda.empty_cache()
        return audio

def create_tts_fn(net_g_ms, hps):
    def tts_fn(text, speaker, sdp_ratio, noise_scale, noise_scale_w, length_scale):
        print(f"{text} | {speaker}")
        sid = hps.data.spk2id[speaker]
        text = text.replace('\n', ' ').replace('\r', '').replace(" ", "")
        if limitation:
            max_len = 100
            if len(text) > max_len:
                return "Error: Text is too long", None
        audio = infer(text, sdp_ratio=sdp_ratio, noise_scale=noise_scale, noise_scale_w=noise_scale_w,
                      length_scale=length_scale, sid=sid, net_g_ms=net_g_ms, hps=hps)
        return "Success", (hps.data.sampling_rate, audio)
    return tts_fn

if __name__ == "__main__":
    device = (
        "cuda:0"
        if torch.cuda.is_available()
        else (
            "mps"
            if sys.platform == "darwin" and torch.backends.mps.is_available()
            else "cpu"
        )
    )

    parser = argparse.ArgumentParser()
    parser.add_argument("--share", default=False, help="make link public", action="store_true")
    parser.add_argument("-d", "--debug", action="store_true", help="enable DEBUG-LEVEL log")
    args = parser.parse_args()
    if args.debug:
        logger.info("Enable DEBUG-LEVEL log")
        logging.basicConfig(level=logging.DEBUG)

    models = []
    with open("pretrained_models/info.json", "r", encoding="utf-8") as f:
        models_info = json.load(f)
    for i, info in models_info.items():
        if not info['enable']:
            continue
        name = info['name']
        title = info['title']
        example = info['example']
        hps = utils.get_hparams_from_file(f"./pretrained_models/{name}/config.json")
        net_g_ms = SynthesizerTrn(
            len(symbols),
            hps.data.filter_length // 2 + 1,
            hps.train.segment_size // hps.data.hop_length,
            n_speakers=hps.data.n_speakers,
            **hps.model)
        utils.load_checkpoint(f'pretrained_models/{i}/{i}.pth', net_g_ms, None, skip_optimizer=True)
        _ = net_g_ms.eval().to(device)
        models.append((name, title, example, list(hps.data.spk2id.keys()), net_g_ms, create_tts_fn(net_g_ms, hps)))
    with gr.Blocks(theme='NoCrypt/miku') as app:
        with gr.Tabs():
            for (name, title, example, speakers, net_g_ms, tts_fn) in models:
                with gr.TabItem(name):
                    with gr.Row():
                        gr.Markdown(
                            '<div align="center">'
                            f'<a><strong>{title}</strong></a>'
                            f'</div>'
                        )
                    with gr.Row():
                        with gr.Column():
                            input_text = gr.Textbox(label="Text (100 words limitation)" if limitation else "Text", lines=5, value=example)
                            btn = gr.Button(value="Generate", variant="primary")
                            with gr.Row():
                                sp = gr.Dropdown(choices=speakers, value=speakers[0], label="Speaker")
                            with gr.Row():
                                sdpr = gr.Slider(label="SDP Ratio", minimum=0, maximum=1, step=0.1, value=0.2)
                                ns = gr.Slider(label="noise_scale", minimum=0.1, maximum=1.0, step=0.1, value=0.6)
                                nsw = gr.Slider(label="noise_scale_w", minimum=0.1, maximum=1.0, step=0.1, value=0.8)
                                ls = gr.Slider(label="length_scale", minimum=0.1, maximum=2.0, step=0.1, value=1)
                        with gr.Column():
                            o1 = gr.Textbox(label="Output Message")
                            o2 = gr.Audio(label="Output Audio")
                        btn.click(tts_fn, inputs=[input_text, sp, sdpr, ns, nsw, ls], outputs=[o1, o2])
        app.queue(concurrency_count=1).launch(share=args.share)