Update app.py

app.py

@@ -1,7 +1,3 @@
-# Step 1: Install Gradio
-#pip install gradio
-
-# Step 2: Import necessary libraries
 import gradio as gr
 import json
 import torch
@@ -10,44 +6,38 @@ from datasets import load_dataset
 import soundfile as sf
 import io
 
-# Step 3: Load the models and the pronunciation dictionary
+# Load models and pronunciation dictionary
 processor = SpeechT5Processor.from_pretrained("microsoft/speecht5_tts")
 model = SpeechT5ForTextToSpeech.from_pretrained("microsoft/speecht5_tts")
 vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan")
 
-# Load pronunciation dictionary from JSON file
 with open("pronunciation_dict.json", "r") as f:
     pronunciation_dict = json.load(f)
 
-# Function to preprocess the input text
+# Preprocess the text
 def preprocess_text(text):
     for term, phonetic in pronunciation_dict.items():
         text = text.replace(term, phonetic)
     return text
 
-# Step 4: Define the TTS function
+# Text-to-Speech function
 def text_to_speech(input_text):
-    # Preprocess the text
     processed_text = preprocess_text(input_text)
-
-    # Convert the processed text to model inputs
     inputs = processor(text=processed_text, return_tensors="pt")
-
-    # Load xvector embeddings from dataset for speaker voice characteristics
+    
     embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
     speaker_embeddings = torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze(0)
-
-    # Generate speech using the model and vocoder
+    
     speech = model.generate_speech(inputs["input_ids"], speaker_embeddings, vocoder=vocoder)
-
-    # Convert generated speech to an in-memory buffer
+    
+    # Prepare audio as BytesIO
     audio_buffer = io.BytesIO()
     sf.write(audio_buffer, speech.numpy(), samplerate=16000, format='WAV')
     audio_buffer.seek(0)
-
+    
     return audio_buffer
 
-# Step 5: Create Gradio interface with examples
+# Define examples
 examples = [
     "We are using API for authentication.",
     "CUDA and TensorFlow work together for deep learning models.",
@@ -56,14 +46,16 @@ examples = [
     "Natural language processing techniques are widely adopted."
 ]
 
+# Create Gradio interface
 iface = gr.Interface(
     fn=text_to_speech,
     inputs="text",
     outputs="audio",
     title="Fine-tuning TTS for English with a Focus on Technical Vocabulary Using SpeechT5",
     description="Enter text with technical jargon for TTS conversion.",
-    examples=examples  # Adding preset examples for users
+    examples=examples,
+    cache_examples=False  # Disable caching for now
 )
 
-# Step 6: Launch the app
-iface.launch(share=True)
+# Launch interface
+iface.launch()