File size: 3,967 Bytes
374f426
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import os
import logging
import sys

import numpy as np

from modules.devices import devices
from modules.synthesize_audio import synthesize_audio
from modules.hf import spaces
from modules.webui import webui_config

logging.basicConfig(
    level=os.getenv("LOG_LEVEL", "INFO"),
    format="%(asctime)s - %(name)s - %(levelname)s - %(message)s",
)


import gradio as gr

import torch

from modules.ssml import parse_ssml
from modules.SynthesizeSegments import SynthesizeSegments, combine_audio_segments

from modules.speaker import speaker_mgr
from modules.data import styles_mgr

from modules.api.utils import calc_spk_style
import modules.generate_audio as generate

from modules.normalization import text_normalize
from modules import refiner, config

from modules.utils import env, audio
from modules.SentenceSplitter import SentenceSplitter


def get_speakers():
    return speaker_mgr.list_speakers()


def get_styles():
    return styles_mgr.list_items()


def segments_length_limit(segments, total_max: int):
    ret_segments = []
    total_len = 0
    for seg in segments:
        if "text" not in seg:
            continue
        total_len += len(seg["text"])
        if total_len > total_max:
            break
        ret_segments.append(seg)
    return ret_segments


@torch.inference_mode()
@spaces.GPU
def synthesize_ssml(ssml: str, batch_size=4):
    try:
        batch_size = int(batch_size)
    except Exception:
        batch_size = 8

    ssml = ssml.strip()

    if ssml == "":
        return None

    segments = parse_ssml(ssml)
    max_len = webui_config.ssml_max
    segments = segments_length_limit(segments, max_len)

    if len(segments) == 0:
        return None

    synthesize = SynthesizeSegments(batch_size=batch_size)
    audio_segments = synthesize.synthesize_segments(segments)
    combined_audio = combine_audio_segments(audio_segments)

    return audio.pydub_to_np(combined_audio)


@torch.inference_mode()
@spaces.GPU
def tts_generate(
    text,
    temperature,
    top_p,
    top_k,
    spk,
    infer_seed,
    use_decoder,
    prompt1,
    prompt2,
    prefix,
    style,
    disable_normalize=False,
    batch_size=4,
):
    try:
        batch_size = int(batch_size)
    except Exception:
        batch_size = 4

    max_len = webui_config.tts_max
    text = text.strip()[0:max_len]

    if text == "":
        return None

    if style == "*auto":
        style = None

    if isinstance(top_k, float):
        top_k = int(top_k)

    params = calc_spk_style(spk=spk, style=style)
    spk = params.get("spk", spk)

    infer_seed = infer_seed or params.get("seed", infer_seed)
    temperature = temperature or params.get("temperature", temperature)
    prefix = prefix or params.get("prefix", prefix)
    prompt1 = prompt1 or params.get("prompt1", "")
    prompt2 = prompt2 or params.get("prompt2", "")

    infer_seed = np.clip(infer_seed, -1, 2**32 - 1, out=None, dtype=np.int64)
    infer_seed = int(infer_seed)

    if not disable_normalize:
        text = text_normalize(text)

    sample_rate, audio_data = synthesize_audio(
        text=text,
        temperature=temperature,
        top_P=top_p,
        top_K=top_k,
        spk=spk,
        infer_seed=infer_seed,
        use_decoder=use_decoder,
        prompt1=prompt1,
        prompt2=prompt2,
        prefix=prefix,
        batch_size=batch_size,
    )

    audio_data = audio.audio_to_int16(audio_data)
    return sample_rate, audio_data


@torch.inference_mode()
@spaces.GPU
def refine_text(text: str, prompt: str):
    text = text_normalize(text)
    return refiner.refine_text(text, prompt=prompt)


@torch.inference_mode()
@spaces.GPU
def split_long_text(long_text_input):
    spliter = SentenceSplitter(webui_config.spliter_threshold)
    sentences = spliter.parse(long_text_input)
    sentences = [text_normalize(s) for s in sentences]
    data = []
    for i, text in enumerate(sentences):
        data.append([i, text, len(text)])
    return data