Update app.py

app.py

@@ -2,6 +2,7 @@ import gradio as gr
 import torch
 from TTS.api import TTS
 import os
+import librosa
 
 # Get device
 device = "cuda" if torch.cuda.is_available() else "cpu"
@@ -9,18 +10,39 @@ device = "cuda" if torch.cuda.is_available() else "cpu"
 # Initialize TTS model
 tts = TTS(model_name="voice_conversion_models/multilingual/vctk/freevc24", progress_bar=False).to(device)
 
-# Get examples from Examples folder
-examples_folder = "Examples/"
-example_files = [f for f in os.listdir(examples_folder) if f.endswith(".wav")]
+def convert_audio_to_wav(file_path):
+    """Convert any supported format (mp3, etc.) to wav using librosa"""
+    output_path = "temp_input.wav"
+    audio, sr = librosa.load(file_path, sr=None)  # Load file (wav, mp3, etc.)
+    librosa.output.write_wav(output_path, audio, sr)  # Convert to wav
+    return output_path
 
-def voice_conversion(input_audio, target_voice):
+def voice_conversion(input_audio, target_voice, uploaded_target_voice):
     output_path = "output.wav"
+    
+    # Check if the user uploaded a target voice, otherwise use selected from examples
+    if uploaded_target_voice is not None:
+        target_voice_path = uploaded_target_voice
+        # Convert uploaded target to wav if necessary
+        if not uploaded_target_voice.endswith(".wav"):
+            target_voice_path = convert_audio_to_wav(uploaded_target_voice)
+    else:
+        target_voice_path = os.path.join("Examples", target_voice)
+        if not os.path.exists(target_voice_path):
+            return "Error: Target voice file not found."
+    
+    # Convert input audio to wav if necessary
+    if not input_audio.endswith(".wav"):
+        input_audio = convert_audio_to_wav(input_audio)
+    
     # Perform voice conversion
-    target_voice = f"{examples_folder}{target_voice}"
-    print(f"Target voice is: {target_voice}")
-    tts.voice_conversion_to_file(source_wav=input_audio, target_wav=target_voice, file_path=output_path)
+    tts.voice_conversion_to_file(source_wav=input_audio, target_wav=target_voice_path, file_path=output_path)
     return output_path
 
+# Get examples from Examples folder
+examples_folder = "Examples/"
+example_files = [f for f in os.listdir(examples_folder) if f.endswith(".wav")]
+
 # Define Gradio Interface
 with gr.Blocks() as demo:
     gr.Markdown("## Voice Conversion using Coqui TTS")
@@ -29,11 +51,24 @@ with gr.Blocks() as demo:
         input_audio = gr.Audio(label="Record or Upload Your Voice", type="filepath")
         target_voice = gr.Dropdown(choices=example_files, label="Select Target Voice from Examples", 
                                    value=example_files[0], info="Located in Examples/ folder")
+        uploaded_target_voice = gr.Audio(label="Or Upload Your Own Target Voice (Overrides Examples)", type="filepath", optional=True)
 
+    with gr.Row():
+        play_button = gr.Button("Preview Selected Target Voice")
+        preview_audio = gr.Audio(label="Preview Target Voice", type="filepath")
+    
+    # Add convert button and output audio
     convert_button = gr.Button("Convert Voice")
     output_audio = gr.Audio(label="Converted Voice", type="filepath")
-
-    convert_button.click(voice_conversion, inputs=[input_audio, target_voice], outputs=output_audio)
+    
+    # Preview button for listening to the selected target voice from examples
+    def preview_target_voice(selected_target_voice):
+        return os.path.join(examples_folder, selected_target_voice)
+    
+    play_button.click(preview_target_voice, inputs=[target_voice], outputs=preview_audio)
+    
+    # Conversion process
+    convert_button.click(voice_conversion, inputs=[input_audio, target_voice, uploaded_target_voice], outputs=output_audio)
 
 # Launch with public=True for public URL access and share link
 demo.launch(share=True)