File size: 3,328 Bytes
19bdfd0
00f303b
 
cdd5e99
19bdfd0
 
00f303b
00b0d1a
3499016
00b0d1a
 
 
 
 
19bdfd0
00b0d1a
19bdfd0
00b0d1a
3499016
18d8a13
7ce66ad
00b0d1a
19bdfd0
00b0d1a
 
cdd5e99
 
 
 
 
 
 
 
 
 
00b0d1a
 
 
cdd5e99
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
00b0d1a
 
 
 
cdd5e99
 
 
00b0d1a
 
 
 
 
 
 
 
3818b2f
 
 
cdd5e99
 
 
 
 
 
 
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
import streamlit as st
import numpy as np
import torch
from transformers import SpeechT5Processor, SpeechT5ForTextToSpeech, SpeechT5HifiGan
from io import StringIO
import soundfile as sf

# Load models outside of function calls for efficiency
@st.cache_data
def load_models():
    model = SpeechT5ForTextToSpeech.from_pretrained("microsoft/speecht5_tts")
    processor = SpeechT5Processor.from_pretrained("microsoft/speecht5_tts")
    vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan")
    return model, processor, vocoder

model, processor, vocoder = load_models()

# Load speaker embeddings
@st.cache_data
def get_speaker_embeddings():
    speaker_embeddings = np.load("cmu_us_clb_arctic-wav-arctic_a0144.npy")
    return torch.tensor(speaker_embeddings).unsqueeze(0)

speaker_embeddings = get_speaker_embeddings()

# Improved Styling
def local_css(file_name):
    with open(file_name) as f:
        st.markdown(f'<style>{f.read()}</style>', unsafe_allow_html=True)

local_css("style.css")

# Streamlined Layout
st.title("Text-to-Voice Conversion")
st.markdown("Convert your text to speech using advanced AI models.")

# Function to convert text to speech
def text_to_speech(text):
    try:
        # Segment the text if it's too long
        max_length = 100  # Set a max length as per model's capability
        segments = [text[i:i+max_length] for i in range(0, len(text), max_length)]
        audio_paths = []

        for segment in segments:
            inputs = processor(text=segment, return_tensors="pt")
            spectrogram = model.generate_speech(inputs["input_ids"], speaker_embeddings)
            with torch.no_grad():
                speech = vocoder(spectrogram)
                audio_path = f"speech_segment_{len(audio_paths)}.wav"
                sf.write(audio_path, speech.numpy(), samplerate=16000)
                audio_paths.append(audio_path)

        return audio_paths
    except Exception as e:
        st.error(f"Error in text-to-speech conversion: {e}")
        return []

# Function to combine audio segments
def combine_audio_segments(paths):
    combined_speech = []
    for path in paths:
        data, samplerate = sf.read(path)
        combined_speech.extend(data)
    sf.write("combined_speech.wav", np.array(combined_speech), samplerate)
    return "combined_speech.wav"

# Text Input
text = st.text_area("Type your text or upload a text file below.")

# Convert Button
if st.button("Convert"):
    if text:
        audio_paths = text_to_speech(text)
        combined_audio_path = combine_audio_segments(audio_paths)
        audio_file = open(combined_audio_path, 'rb')
        audio_bytes = audio_file.read()
        st.audio(audio_bytes, format='audio/wav')
    else:
        st.error("Please enter some text to convert.")

# File Uploader
uploaded_file = st.file_uploader("Upload your text file here", type=['txt'])
if uploaded_file is not None:
    stringio = StringIO(uploaded_file.getvalue().decode("utf-8"))
    text = stringio.read()
    st.write(text)

    if st.button("Convert Uploaded File", key=1):
        audio_paths = text_to_speech(text)
        combined_audio_path = combine_audio_segments(audio_paths)
        audio_file = open(combined_audio_path, 'rb')
        audio_bytes = audio_file.read()
        st.audio(audio_bytes, format='audio/wav')