File size: 1,573 Bytes
eefe9ae
 
2ec67ac
 
eefe9ae
 
 
ebb2d0b
 
 
 
 
564910c
ebb2d0b
564910c
 
ebb2d0b
 
 
 
 
 
 
 
 
 
 
564910c
ebb2d0b
eefe9ae
564910c
2ec67ac
eefe9ae
 
 
564910c
 
 
 
eefe9ae
 
 
 
564910c
 
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
import gradio as gr
from gtts import gTTS
from pydub import AudioSegment
import numpy as np
import os

def text_to_speech(prompt):
    # gTTS๋ฅผ ์ด์šฉํ•ด Bulgarian ํ…์ŠคํŠธ๋ฅผ ์Œ์„ฑ์œผ๋กœ ๋ณ€ํ™˜
    tts = gTTS(text=prompt, lang="bg")
    audio_file = "output.mp3"
    tts.save(audio_file)
    
    # pydub๋ฅผ ์‚ฌ์šฉํ•˜์—ฌ mp3 ํŒŒ์ผ์„ ๋ถˆ๋Ÿฌ์˜ต๋‹ˆ๋‹ค.
    sound = AudioSegment.from_mp3(audio_file)
    
    # pydub์˜ raw ๋ฐ์ดํ„ฐ๋ฅผ numpy ๋ฐฐ์—ด๋กœ ๋ณ€ํ™˜ (int16)
    samples = np.array(sound.get_array_of_samples())
    
    # ๋งŒ์•ฝ ์Šคํ…Œ๋ ˆ์˜ค๋ผ๋ฉด ๋ชจ๋…ธ๋กœ ๋ณ€ํ™˜ (์ฑ„๋„ ํ‰๊ท )
    if sound.channels > 1:
        samples = samples.reshape((-1, sound.channels))
        samples = samples.mean(axis=1)
    
    # int16 ๋ฐ์ดํ„ฐ๋ฅผ float32๋กœ ์ •๊ทœํ™” (๋ฒ”์œ„: [-1.0, 1.0])
    samples = samples.astype(np.float32) / 32768.0
    sample_rate = sound.frame_rate

    # ์ž„์‹œ๋กœ ์ƒ์„ฑํ•œ mp3 ํŒŒ์ผ ์‚ญ์ œ
    os.remove(audio_file)
    
    # gr.Audio(type="numpy")๋Š” (numpy_array, sample_rate) ํŠœํ”Œ์„ ๊ธฐ๋Œ€ํ•ฉ๋‹ˆ๋‹ค.
    return samples, sample_rate

with gr.Blocks() as demo:
    gr.Markdown("## Bulgarian Text-to-Speech (TTS)")
    with gr.Row():
        input_prompt = gr.Textbox(label="Enter a prompt in Bulgarian:")
        # type์„ "numpy"๋กœ ์„ค์ •ํ•˜์—ฌ numpy ๋ฐฐ์—ด์„ ์‚ฌ์šฉํ•ฉ๋‹ˆ๋‹ค.
        output_audio = gr.Audio(label="Generated Speech", type="numpy")
    generate_button = gr.Button("Generate Speech")
    
    generate_button.click(text_to_speech, inputs=input_prompt, outputs=output_audio)

if __name__ == "__main__":
    demo.launch()