File size: 3,334 Bytes
19b10bb 0812f72 19b10bb 09778ba b80857a fa86cdd 52b095f 80d1ea2 19b10bb b98f522 bf20a09 b98f522 19b10bb b80857a 19b10bb e38db12 19b10bb 27c1d79 f4c0261 e38db12 19b10bb bf20a09 19b10bb 09778ba 19b10bb 09778ba 19b10bb 80d1ea2 09778ba 19b10bb 09778ba 27c1d79 |
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 |
import torch
from transformers import SpeechT5ForTextToSpeech, SpeechT5Processor, SpeechT5HifiGan
import soundfile as sf
import gradio as gr
import scipy.io.wavfile as wav
import numpy as np
import wave
from datasets import load_dataset, Audio, config
from IPython.display import Audio
# Load the TTS model from the Hugging Face Hub
checkpoint = "arham061/speecht5_finetuned_voxpopuli_nl" # Replace with your actual model name
processor = SpeechT5Processor.from_pretrained(checkpoint)
model = SpeechT5ForTextToSpeech.from_pretrained(checkpoint)
tokenizer = processor.tokenizer
vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan")
# Buckwalter to Unicode mapping
buck2uni = {
u"\u0627":"A",
u"\u0627":"A",
u"\u0675":"A",
u"\u0673":"A",
u"\u0630":"A",
u"\u0622":"AA",
u"\u0628":"B",
u"\u067E":"P",
u"\u062A":"T",
u"\u0637":"T",
u"\u0679":"T",
u"\u062C":"J",
u"\u0633":"S",
u"\u062B":"S",
u"\u0635":"S",
u"\u0686":"CH",
u"\u062D":"H",
u"\u0647":"H",
u"\u0629":"H",
u"\u06DF":"H",
u"\u062E":"KH",
u"\u062F":"D",
u"\u0688":"D",
u"\u0630":"Z",
u"\u0632":"Z",
u"\u0636":"Z",
u"\u0638":"Z",
u"\u068E":"Z",
u"\u0631":"R",
u"\u0691":"R",
u"\u0634":"SH",
u"\u063A":"GH",
u"\u0641":"F",
u"\u06A9":"K",
u"\u0642":"K",
u"\u06AF":"G",
u"\u0644":"L",
u"\u0645":"M",
u"\u0646":"N",
u"\u06BA":"N",
u"\u0648":"O",
u"\u0649":"Y",
u"\u0626":"Y",
u"\u06CC":"Y",
u"\u06D2":"E",
u"\u06C1":"H",
u"\u064A":"E" ,
u"\u06C2":"AH" ,
u"\u06BE":"H" ,
u"\u0639":"A" ,
u"\u0643":"K" ,
u"\u0621":"A",
u"\u0624":"O",
u"\u060C":"" #seperator ulta comma
}
def transString(string, reverse=0):
"""Given a Unicode string, transliterate into Buckwalter. To go from
Buckwalter back to Unicode, set reverse=1"""
for k, v in buck2uni.items():
if not reverse:
string = string.replace(k, v)
else:
string = string.replace(v, k)
return string
def generate_audio(text):
# Convert input text to Roman Urdu
roman_urdu = transString(text)
# Tokenize the input text
inputs = processor(text=roman_urdu, return_tensors="pt")
# Generate audio from the SpeechT5 model
speaker_embeddings = torch.tensor(np.load("speaker_embeddings.npy"))
speech = model.generate_speech(inputs["input_ids"], speaker_embeddings, vocoder=vocoder)
return speech
def text_to_speech(text):
# Generate audio
audio_output = generate_audio(text)
return Audio(audio_output.numpy(), rate=16000)
# Define the Gradio interface
inputs = gr.inputs.Textbox(label="Enter text in Urdu")
outputs = gr.outputs.Audio(label="Audio")
interface = gr.Interface(fn=text_to_speech, inputs=inputs, outputs=outputs, title="Urdu TTS")
interface.launch()
|