Update app.py

app.py

@@ -1,6 +1,7 @@
 import gradio as gr
 import moviepy.editor as mp
 import librosa
+import numpy as np
 from transformers import pipeline
 
 # Load Whisper model for speech-to-text
@@ -45,20 +46,34 @@ def generate_subtitles(video_file, language_name):
         # Load the audio file as a waveform using librosa
         waveform, sr = librosa.load(audio_path, sr=16000)  # sr=16000 for Whisper
 
-        # Pass the waveform (NumPy array) directly to Whisper's ASR model
-        transcription = asr(waveform)["text"]
+        # Process audio in chunks
+        chunk_duration = 30  # seconds
+        chunk_size = sr * chunk_duration  # number of samples per chunk
+        transcriptions = []
+
+        for i in range(0, len(waveform), chunk_size):
+            chunk = waveform[i:i + chunk_size]
+            if len(chunk) == 0:
+                break  # Avoid processing empty chunks
+            
+            # Pass the chunk to Whisper's ASR model
+            transcription = asr(chunk)["text"]
+            transcriptions.append(transcription)
+
+        # Combine all transcriptions into a single string
+        full_transcription = " ".join(transcriptions)
 
         print("Starting translation")
 
         # Translate transcription to the target language using M2M100
         translation_pipeline = pipeline('translation', model='facebook/m2m100_418M')
         translated_subtitles = translation_pipeline(
-            transcription, 
+            full_transcription, 
             forced_bos_token_id=translation_pipeline.tokenizer.get_lang_id(target_language)
         )[0]["translation_text"]
 
         # Return subtitles
-        subtitles = f"Original: {transcription}\nTranslated: {translated_subtitles}"
+        subtitles = f"Original: {full_transcription}\nTranslated: {translated_subtitles}"
         return subtitles
     
     except Exception as e: