File size: 4,743 Bytes
fca9b48
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
209c75f
fca9b48
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
209c75f
 
fca9b48
 
 
209c75f
 
fca9b48
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
320bb9e
fca9b48
 
 
 
 
 
320bb9e
fca9b48
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bf0afab
fca9b48
 
 
 
 
 
 
 
 
 
 
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
import os
import torch
import gradio as gr
import torchaudio
import time
from datetime import datetime
from tortoise.api import TextToSpeech
from tortoise.utils.text import split_and_recombine_text
from tortoise.utils.audio import load_audio, load_voice, load_voices

VOICE_OPTIONS = [
    "angie",
    "cond_latent_example",
    "deniro",
    "freeman",
    "halle",
    "lj",
    "myself",
    "pat2",
    "snakes",
    "tom",
    "train_daws",
    "train_dreams",
    "train_grace",
    "train_lescault",
    "weaver",
    "applejack",
    "daniel",
    "emma",
    "geralt",
    "jlaw",
    "mol",
    "pat",
    "rainbow",
    "tim_reynolds",
    "train_atkins",
    "train_dotrice",
    "train_empire",
    "train_kennard",
    "train_mouse",
    "william",
    "random",  # special option for random voice
    "disabled",  # special option for disabled voice
]


def inference(
    text,
    script,
    voice,
    voice_b,
    preset,
    seed,
    regenerate,
    split_by_newline,
):
    if regenerate.strip() == "":
        regenerate = None
    name = "generated.wav"
    if text is None or text.strip() == "":
        with open(script.name) as f:
            text = f.read()
        if text.strip() == "":
            raise gr.Error("Please provide either text or script file with content.")

    if split_by_newline == "Yes":
        texts = list(filter(lambda x: x.strip() != "", text.split("\n")))
    else:
        texts = split_and_recombine_text(text)

    if regenerate is not None:
        regenerate = list(map(int, regenerate.split()))

    voices = [voice]
    if voice_b != "disabled":
        voices.append(voice_b)

    if len(voices) == 1:
        voice_samples, conditioning_latents = load_voice(voice)
    else:
        voice_samples, conditioning_latents = load_voices(voices)

    start_time = time.time()

    all_parts = []
    for j, text in enumerate(texts):
        gen = tts.tts_with_preset(
            text,
            voice_samples=voice_samples,
            conditioning_latents=conditioning_latents,
            preset=preset,
            k=1,
            use_deterministic_seed=seed,
        )

        audio_ = gen.squeeze(0).cpu()
        all_parts.append(audio_)

    full_audio = torch.cat(all_parts, dim=-1)

    # os.makedirs("outputs", exist_ok=True)
    # torchaudio.save(os.path.join("outputs", f"{name}.wav"), full_audio, 24000)

    with open("Tortoise_TTS_Runs_Scripts.log", "a") as f:
        f.write(
            f"{datetime.now()} | Voice: {','.join(voices)} | Text: {text} | Quality: {preset} | Time Taken (s): {time.time()-start_time} | Seed: {seed}\n"
        )

    output_texts = [f"({j+1}) {texts[j]}" for j in range(len(texts))]

    return ((24000, full_audio.squeeze().cpu().numpy()), "\n".join(output_texts))


def main():
    text = gr.Textbox(
        lines=4,
        label="Text (Provide either text, or upload a newline separated text file below):",
    )
    script = gr.File(label="Upload a text file")
    name = gr.Textbox(
        lines=1, label="Name of the output file / folder to store intermediate results:"
    )
    preset = gr.Radio(
        ["ultra_fast", "fast", "standard", "high_quality"],
        value="fast",
        label="Preset mode (determines quality with tradeoff over speed):",
        type="value",
    )
    voice = gr.Dropdown(
        VOICE_OPTIONS, value="angie", label="Select voice:", type="value"
    )
    voice_b = gr.Dropdown(
        VOICE_OPTIONS,
        value="disabled",
        label="(Optional) Select second voice:",
        type="value",
    )
    seed = gr.Number(value=0, precision=0, label="Seed (for reproducibility):")
    regenerate = gr.Textbox(
        lines=1,
        label="Comma-separated indices of clips to regenerate [starting from 1]",
    )

    split_by_newline = gr.Radio(
        ["Yes", "No"],
        label="Split by newline (If [No], it will automatically try to find relevant splits):",
        type="value",
        value="No",
    )

    output_audio = gr.Audio(label="Combined audio:")
    output_text = gr.Textbox(label="Split texts with indices:", lines=10)

    interface = gr.Interface(
        fn=inference,
        inputs=[
            text,
            script,
            voice,
            voice_b,
            preset,
            seed,
            regenerate,
            split_by_newline,
        ],
        outputs=[output_audio, output_text],
    )
    interface.launch()


if __name__ == "__main__":
    tts = TextToSpeech(kv_cache=True, use_deepspeed=True, half=True)

    with open("Tortoise_TTS_Runs_Scripts.log", "a") as f:
        f.write(
            f"\n\n-------------------------Tortoise TTS Scripts Logs, {datetime.now()}-------------------------\n"
        )

    main()