Update app.py

app.py

@@ -1,14 +1,17 @@
 import streamlit as st
 from moviepy.editor import VideoFileClip
 import whisper
-from googletrans import Translator
+from translate import Translator
 from gtts import gTTS
 import tempfile
 import os
 import numpy as np
 
 # Initialize Whisper model
-whisper_model = whisper.load_model("base")
+try:
+    whisper_model = whisper.load_model("base")  # Ensure the model is installed from the correct Whisper library
+except AttributeError:
+    st.error("Whisper model could not be loaded. Ensure that Whisper is installed from GitHub.")
 
 # Language options
 LANGUAGES = {
@@ -44,22 +47,28 @@ if video_file:
             os.remove(temp_video_path)
             st.stop()
 
-        # Transcribe audio using Whisper in chunks
-        try:
-            # Load the audio file with Whisper
-            audio = whisper.load_audio(audio_path)
-            audio_segments = split_audio(audio, segment_length=30)  # Split into 30-second segments
+        # Function to transcribe audio in chunks
+        def transcribe_audio_in_chunks(audio_path, model, chunk_length=30):
+            audio_clip = whisper.load_audio(audio_path)
+            audio_duration = whisper.get_duration(audio_clip)
+            segments = []
+
+            for start in np.arange(0, audio_duration, chunk_length):
+                end = min(start + chunk_length, audio_duration)
+                segment = audio_clip[int(start * 16000):int(end * 16000)]  # Convert to the right format
+                result = model.transcribe(segment)
+                segments.append(result['text'])
 
-            original_text = ""
-            for segment in audio_segments:
-                result = whisper_model.transcribe(segment)
-                original_text += result["text"] + " "  # Concatenate transcriptions
+            return ' '.join(segments)
 
-            st.write("Original Transcription:", original_text.strip())
+        # Transcribe audio using Whisper
+        try:
+            original_text = transcribe_audio_in_chunks(audio_path, whisper_model)
+            st.write("Original Transcription:", original_text)
 
             # Translate text to the target language
-            translator = Translator()
-            translated_text = translator.translate(original_text.strip(), dest=LANGUAGES[target_language]).text
+            translator = Translator(to_lang=LANGUAGES[target_language])
+            translated_text = translator.translate(original_text)
             st.write(f"Translated Text ({target_language}):", translated_text)
 
             # Convert translated text to speech
@@ -77,13 +86,3 @@ if video_file:
         os.remove(temp_video_path)
         os.remove(audio_path)
         os.remove(audio_output_path)
-
-def split_audio(audio, segment_length=30):
-    """Split audio into segments of specified length in seconds."""
-    total_length = audio.shape[1]  # Total length in seconds
-    segments = []
-    for start in range(0, total_length, segment_length):
-        end = min(start + segment_length, total_length)
-        segment = audio[:, start:end]  # Append the segment
-        segments.append(segment)
-    return segments