Update app.py

app.py

@@ -1,15 +1,22 @@
 import streamlit as st
-import numpy as np
 import os
-import soundfile as sf
 from TTS.api import TTS
-import torch
+import moviepy.editor as mp
 from io import BytesIO
+import soundfile as sf
 
 # Set up the model for text-to-speech (TTS)
 MODEL_NAME = "tts_models/en/ljspeech/tacotron2-DDC"  # Example TTS model; adjust as needed
 tts = TTS(model_name=MODEL_NAME, progress_bar=True, gpu=False)
 
+# Function to extract audio from MP4 file
+def extract_audio_from_mp4(mp4_file):
+    video = mp.VideoFileClip(mp4_file)
+    audio = video.audio
+    audio_file = "temp_audio.wav"
+    audio.write_audiofile(audio_file)
+    return audio_file
+
 # Function to load audio file
 def load_audio(file):
     audio_data, sample_rate = sf.read(file)
@@ -24,23 +31,40 @@ def save_audio(output_audio, sample_rate):
 # Streamlit app
 def main():
     st.title("Voice Cloning Tool")
-    st.markdown("Upload a voice input, and get the cloned voice output.")
+    st.markdown("Upload an MP4, WAV, or MP3 file, and get the cloned voice output.")
 
     # File upload
-    audio_file = st.file_uploader("Upload your audio file", type=["wav", "mp3","mp4"])
-    
+    audio_file = st.file_uploader("Upload your audio file", type=["wav", "mp3", "mp4"])
+
     if audio_file is not None:
-        st.audio(audio_file, format="audio/wav")
-        
-        # Load audio file
-        audio_data, sample_rate = load_audio(audio_file)
+        # Handle MP4 file by extracting audio
+        if audio_file.type == "video/mp4":
+            # Save the uploaded file to a temporary location
+            with open("uploaded_video.mp4", "wb") as f:
+                f.write(audio_file.getbuffer())
+
+            # Extract audio from MP4
+            audio_path = extract_audio_from_mp4("uploaded_video.mp4")
+            st.audio(audio_path, format="audio/wav")
+            
+            # Load audio for TTS processing
+            audio_data, sample_rate = load_audio(audio_path)
+        else:
+            # For audio files directly
+            st.audio(audio_file, format=f"audio/{audio_file.type.split('/')[1]}")
+
+            # Load audio file
+            with open("temp_audio.wav", "wb") as f:
+                f.write(audio_file.getbuffer())
+            
+            audio_data, sample_rate = load_audio("temp_audio.wav")
 
         # Perform voice cloning (This assumes your TTS model supports some form of input)
         try:
             st.text("Processing your input...")
             output_audio = tts.tts(audio_data)  # Pass the audio to your TTS model for cloning
             output_path = save_audio(output_audio, sample_rate)
-            
+
             # Provide download link
             st.audio(output_path, format="audio/wav")
             st.markdown(f"[Download Cloned Voice](/{output_path})")