File size: 2,672 Bytes
e4eb5c5
944dedf
 
 
 
4492d6d
19be65d
a82f51b
 
 
 
 
 
 
4fda610
 
a82f51b
 
 
 
 
 
 
 
 
 
4fda610
a82f51b
 
 
 
 
 
 
 
 
 
 
 
 
 
4eb15f6
944dedf
8ce2dae
19be65d
7e5c84b
686e3d3
944dedf
 
f6a94c1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
944dedf
1782e10
 
 
 
4e24051
f6a94c1
1782e10
 
4e24051
f845b05
944dedf
f6a94c1
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
import gradio as gr
import wave
import numpy as np
from io import BytesIO
from huggingface_hub import hf_hub_download
from piper import PiperVoice 
from transformers import pipeline
import hazm
import typing

normalizer = hazm.Normalizer()
sent_tokenizer = hazm.SentenceTokenizer()
word_tokenizer = hazm.WordTokenizer()

tagger_path = hf_hub_download(repo_id="gyroing/HAZM_POS_TAGGER", filename="pos_tagger.model")
tagger = hazm.POSTagger(model=tagger_path)

def preprocess_text(text: str) -> typing.List[typing.List[str]]:
        """Split/normalize text into sentences/words with hazm"""
        text = normalizer.normalize(text)
        processed_sentences = []

        for sentence in sent_tokenizer.tokenize(text):
            words = word_tokenizer.tokenize(sentence)
            processed_words = fix_words(words)
            processed_sentences.append(" ".join(processed_words))
        return  " ".join(processed_sentences)    
def fix_words(words: typing.List[str]) -> typing.List[str]:
        fixed_words = []

        for word, pos in tagger.tag(words):
            if pos[-1] == "Z":
                if word[-1] != "ِ":
                    if (word[-1] == "ه") and (word[-2] != "ا"):
                        word += "‌ی"
                word += "ِ"
                    

            fixed_words.append(word)

        return fixed_words

def synthesize_speech(text):


    model_path = hf_hub_download(repo_id="gyroing/Persian-Piper-Model-gyro", filename="fa_IR-gyro-meduim.onnx")
    config_path = hf_hub_download(repo_id="gyroing/Persian-Piper-Model-gyro", filename="fa_IR-gyro-meduim.onnx.json")
    voice = PiperVoice.load(model_path, config_path)

    # Create an in-memory buffer for the WAV file
    buffer = BytesIO()
    with wave.open(buffer, 'wb') as wav_file:
        wav_file.setframerate(voice.config.sample_rate)
        wav_file.setsampwidth(2)  # 16-bit
        wav_file.setnchannels(1)  # mono

        # Synthesize speech
        voice.synthesize(text, wav_file)

    # Convert buffer to NumPy array for Gradio output
    buffer.seek(0)
    audio_data = np.frombuffer(buffer.read(), dtype=np.int16)

    return audio_data.tobytes(), None

# Using Gradio Blocks
with gr.Blocks(theme=gr.themes.Base()) as blocks:
    gr.Markdown("# Text to Speech Synthesizer")
    gr.Markdown("Enter text to synthesize it into speech using PiperVoice.")
    input_text = preprocess_text(gr.Textbox(label="Input Text"))
    output_audio = gr.Audio(label="Synthesized Speech", type="numpy")
    submit_button = gr.Button("Synthesize")

    submit_button.click(synthesize_speech, inputs=input_text, outputs=[output_audio])

# Run the app
blocks.launch()