File size: 3,191 Bytes
ad2ec01
 
 
 
 
 
048d6c5
ad2ec01
 
 
048d6c5
597e9d7
9667c43
048d6c5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ad2ec01
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
048d6c5
 
 
 
 
 
 
 
 
 
 
 
ad2ec01
048d6c5
 
 
 
ad2ec01
138a10c
ad2ec01
 
 
 
 
 
 
 
 
 
 
048d6c5
ad2ec01
 
 
 
 
 
 
 
 
 
 
048d6c5
 
ad2ec01
 
048d6c5
ad2ec01
 
 
 
 
 
 
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
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 = [
    "Kasuri",
    "A",
    "angie",
    "deniro",
    "freeman",
    "halle",
    "lj",
    "myself",
    "pat2",
    "snakes",
    "tom",
    "daws",
    "dreams",
    "grace",
    "lescault",
    "weaver",
    "applejack",
    "daniel",
    "emma",
    "geralt",
    "jlaw",
    "mol",
    "pat",
    "rainbow",
    "tim_reynolds",
    "atkins",
    "dortice",
    "empire",
    "kennard",
    "mouse",
    "william",
    "jane_eyre",
    "random",  # special option for random voice
]


def inference(
    text,
    script,
    voice,
    voice_b,
    seed,
    split_by_newline,
):
    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)

    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):
        for audio_frame in tts.tts_with_preset(
            text,
            voice_samples=voice_samples,
            conditioning_latents=conditioning_latents,
            preset="ultra_fast",
            k=1
        ):
            # print("Time taken: ", time.time() - start_time)
            # all_parts.append(audio_frame)
            yield (24000, audio_frame.cpu().detach().numpy())

    # wav = torch.cat(all_parts, dim=0).unsqueeze(0)
    # print(wav.shape)
    # torchaudio.save("output.wav", wav.cpu(), 24000)
    # yield (None, gr.make_waveform(audio="output.wav",))
def main():
    title = "Tortoise TTS"
    description = """
    """
    text = gr.Textbox(
        lines=4,
        label="Text (Provide either text, or upload a newline separated text file below):",
    )

    voice = gr.Dropdown(
        VOICE_OPTIONS, value="jane_eyre", label="Select voice:", type="value"
    )

    output_audio = gr.Audio(label="streaming audio:", streaming=True, autoplay=True)
    # download_audio = gr.Audio(label="dowanload audio:")
    interface = gr.Interface(
        fn=inference,
        inputs=[
            text,
            voice,
        ],
        title=title,
        description=description,
        outputs=[output_audio],
    )
    interface.queue().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()