Update app.py

app.py

@@ -20,33 +20,10 @@ def adjust_speed(audio_path, speed_factor):
     sf.write(audio_path, y_speeded, sr)
     
 
-"""Set up the Gradio interface"""
-
+# Set up the Gradio interface
 import gradio as gr
-from transformers import pipeline
-from TTS.api import TTS
-
-# Load pre-trained emotion detection model
-emotion_classifier = pipeline("text-classification", model="bhadresh-savani/distilbert-base-uncased-emotion")
-
-# Load TTS model
-tts_model = TTS(model_name="tts_models/en/ljspeech/tacotron2-DDC")
 
-# Emotion-specific settings for pitch and speed
-emotion_settings = {
-    "neutral": {"pitch": 1.0, "speed": 1.0},
-    "joy": {"pitch": 1.3, "speed": 1.2},
-    "sadness": {"pitch": 0.8, "speed": 0.9},
-    "anger": {"pitch": 1.6, "speed": 1.4},
-    "fear": {"pitch": 1.2, "speed": 0.95},
-    "surprise": {"pitch": 1.5, "speed": 1.3},
-    "disgust": {"pitch": 0.9, "speed": 0.95},
-    "shame": {"pitch": 0.8, "speed": 0.85},
-}
-
-
-# Function to process text or file input and generate audio
-def emotion_aware_tts_pipeline(input_text=None, file_input=None):
+def emotion_aware_tts_pipeline_gradio(input_text=None, file_input=None):
     try:
         # Get text from input or file
         if file_input:
@@ -59,19 +36,18 @@ def emotion_aware_tts_pipeline(input_text=None, file_input=None):
             emotion = emotion_data['label']
             confidence = emotion_data['score']
 
-            # Adjust text for speed
+            # Adjust pitch and speed
             settings = emotion_settings.get(emotion.lower(), {"pitch": 1.0, "speed": 1.0})
-            speed = settings["speed"]
             pitch = settings["pitch"]
-
+            speed = settings["speed"]
 
             # Generate audio
             audio_path = "output.wav"
-            tts_model.tts_to_file(text=input_text, file_path=audio_path)
+            mel_spectrogram = tts_model.get_mel_spectrogram(input_text)
+            audio = vocoder.decode(mel_spectrogram)
 
-            # Adjust pitch
-            pitch_factor = (pitch - 1.0) * 12  # Convert to semitones for librosa
-            adjust_pitch(audio_path, pitch_factor)
+            # Post-processing: adjust pitch and speed
+            adjust_pitch_and_speed(audio_path, pitch_factor=pitch, speed_factor=speed)
 
             return f"Detected Emotion: {emotion} (Confidence: {confidence:.2f})", audio_path
         else:
@@ -79,12 +55,9 @@ def emotion_aware_tts_pipeline(input_text=None, file_input=None):
     except Exception as e:
         return f"Error: {str(e)}", None
 
-
-
-
 # Define Gradio interface
-interface = gr.Interface(
-    fn=emotion_aware_tts_pipeline,
+iface = gr.Interface(
+    fn=emotion_aware_tts_pipeline_gradio,
     inputs=[
         gr.Textbox(label="Input Text", placeholder="Enter text here"),
         gr.File(label="Upload a Text File")
@@ -98,21 +71,4 @@ interface = gr.Interface(
 )
 
 # Launch Gradio interface
-interface.launch()# Define Gradio interface
-interface = gr.Interface(
-    fn=emotion_aware_tts_pipeline,
-    inputs=[
-        gr.Textbox(label="Input Text", placeholder="Enter text here"),
-        gr.File(label="Upload a Text File")
-    ],
-    outputs=[
-        gr.Textbox(label="Detected Emotion"),
-        gr.Audio(label="Generated Audio")
-    ],
-    title="Emotion-Aware Text-to-Speech",
-    description="Input text or upload a text file to detect the emotion and generate audio with emotion-aware modulation."
-)
-
-# Export the interface object so Hugging Face can launch it
-if __name__ == "__main__":
-    interface.launch()
+iface.launch()