Create app.py

app.py

@@ -0,0 +1,166 @@
+import gradio as gr
+import torch
+import soundfile as sf
+import spaces
+import os
+import numpy as np
+import re
+from transformers import SpeechT5Processor, SpeechT5ForTextToSpeech, SpeechT5HifiGan
+from speechbrain.pretrained import EncoderClassifier
+from datasets import load_dataset
+
+device = "cuda" if torch.cuda.is_available() else "cpu"
+
+def load_models_and_data():
+    model_name = "microsoft/speecht5_tts"
+    processor = SpeechT5Processor.from_pretrained(model_name)
+    model = SpeechT5ForTextToSpeech.from_pretrained("speecht5_finetuned_Aumkesh_tr").to(device)
+    vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan").to(device)
+    
+    spk_model_name = "speechbrain/spkrec-xvect-voxceleb"
+    speaker_model = EncoderClassifier.from_hparams(
+        source=spk_model_name,
+        run_opts={"device": device},
+        savedir=os.path.join("/tmp", spk_model_name),
+    )
+    
+    # Load a sample from a dataset for default embedding
+    dataset = load_dataset("Yassmen/TTS_English_Technical_data", split="train")
+    example = dataset[304]
+    
+    return model, processor, vocoder, speaker_model, example
+
+model, processor, vocoder, speaker_model, default_example = load_models_and_data()
+
+def create_speaker_embedding(waveform):
+    with torch.no_grad():
+        speaker_embeddings = speaker_model.encode_batch(torch.tensor(waveform).unsqueeze(0).to(device))
+        speaker_embeddings = torch.nn.functional.normalize(speaker_embeddings, dim=2)
+        speaker_embeddings = speaker_embeddings.squeeze()
+    return speaker_embeddings
+
+def prepare_default_embedding(example):
+    audio = example["audio"]
+    return create_speaker_embedding(audio["array"])
+
+default_embedding = prepare_default_embedding(default_example)
+
+replacements = [
+    ('API', 'A-P-I'),
+    ('CUDA', 'Coo-da'),
+    ('ChatGPT', 'Chat-G-P-T'),
+    ('HTTP', 'H-T-T-P'),
+    ('JSON', 'J-S-O-N'),
+    ('GPU', 'G-P-U'),
+    ('RAM', 'R-A-M'),
+    ('CPU', 'C-P-U'),
+    ('SQL', 'S-Q-L'),
+    ('NLP', 'N-L-P'),
+    ('PyTorch', 'Pie-torch'),
+    ('TensorFlow', 'Ten-sor-flow'),
+    ('SaaS', 'SaaS'),
+    ('GitHub', 'Git-Hub'),
+    ('Docker', 'Dock-er'),
+    ('Kubernetes', 'Koo-ber-net-ees'),
+    ('OpenAI', 'Open-A-I'),
+    ('IOT', 'I-O-T'),
+    ('Linux', 'Li-nux'),
+]
+
+number_words = {
+    0: "zero", 1: "one", 2: "two", 3: "tree", 4: "four", 5: "five", 6: "six", 7: "seven", 8: "eight", 9: "nine",
+    10: "ten", 11: "eleven", 12: "twelve", 13: "thirteen", 14: "fourteen", 15: "fifteen", 16: "sixteen", 17: "seventeen",
+    18: "eighteen", 19: "nineteen", 20: "tweenty", 30: "thirty", 40: "forty", 50: "fifty", 60: "sixty", 70: "seventy",
+    80: "eighty", 90: "ninty", 100: "hundred", 1000: "thousand"
+}
+
+def number_to_words(number):
+    if number < 20:
+        return number_words[number]
+    elif number < 100:
+        tens, unit = divmod(number, 10)
+        return number_words[tens * 10] + (" " + number_words[unit] if unit else "")
+    elif number < 1000:
+        hundreds, remainder = divmod(number, 100)
+        return (number_words[hundreds] + " hundred" if hundreds > 1 else " hundred") + (" " + number_to_words(remainder) if remainder else "")
+    elif number < 1000000:
+        thousands, remainder = divmod(number, 1000)
+        return (number_to_words(thousands) + " thousand" if thousands > 1 else " thousand") + (" " + number_to_words(remainder) if remainder else "")
+    elif number < 1000000000:
+        millions, remainder = divmod(number, 1000000)
+        return number_to_words(millions) + " million" + (" " + number_to_words(remainder) if remainder else "")
+    elif number < 1000000000000:
+        billions, remainder = divmod(number, 1000000000)
+        return number_to_words(billions) + " billion" + (" " + number_to_words(remainder) if remainder else "")
+    else:
+        return str(number)
+
+def replace_numbers_with_words(text):
+
+    def replace(match):
+        number = int(match.group())
+        return number_to_words(number)
+
+    # Find the numbers and change with words.
+    result = re.sub(r'\b\d+\b', replace, text)
+
+    return result
+
+def replace_numbers_with_words(text):
+    def replace(match):
+        number = int(match.group())
+        return number_to_words(number)
+
+    # Find the numbers and change with words.
+    result = re.sub(r'\b\d+\b', replace, text)
+
+    return result
+
+def normalize_text(text):
+    # Convert to lowercase
+    text = text.lower()
+    
+    # Replace numbers with words
+    text = replace_numbers_with_words(text)
+    
+    # Apply character replacements
+    for old, new in replacements:
+        text = text.replace(old, new)
+    
+    # Remove punctuation
+    text = re.sub(r'[^\w\s]', '', text)
+    
+    return text
+
+@spaces.GPU(duration=60)
+def text_to_speech(text, audio_file=None):
+    # Normalize the input text
+    normalized_text = normalize_text(text)
+    
+    # Prepare the input for the model
+    inputs = processor(text=normalized_text, return_tensors="pt").to(device)
+    
+    # Use the default speaker embedding
+    speaker_embeddings = default_embedding
+    
+    # Generate speech
+    with torch.no_grad():
+        speech = model.generate_speech(inputs["input_ids"], speaker_embeddings.unsqueeze(0), vocoder=vocoder)
+    
+    speech_np = speech.cpu().numpy()
+
+    return (16000, speech_np)
+    
+iface = gr.Interface(
+    fn=text_to_speech,
+    inputs=[
+        gr.Textbox(label="Enter Turkish text to convert to speech")
+    ],
+    outputs=[
+        gr.Audio(label="Generated Speech", type="numpy")
+    ],
+    title="English SpeechT5 Text-to-Speech Demo",
+    description="Enter English text, and listen to the generated speech."
+)
+
+iface.launch(share=True)