Update app.py

app.py

@@ -31,9 +31,6 @@ pipe = pipeline(
 pipe.model.config.forced_decoder_ids = pipe.tokenizer.get_decoder_prompt_ids(language=language, task="transcribe")
 
 # Similarity check to remove repeated phrases
-def is_similar(a, b, threshold=0.8):
-    return SequenceMatcher(None, a, b).ratio() > threshold
-
 def remove_repeated_phrases(text):
     sentences = re.split(r'(?<=[。！？])', text)
     cleaned_sentences = []
@@ -42,23 +39,22 @@ def remove_repeated_phrases(text):
             cleaned_sentences.append(sentence.strip())
     return " ".join(cleaned_sentences)
 
-# Remove punctuation
 def remove_punctuation(text):
     return re.sub(r'[^\w\s]', '', text)
 
-# Transcription function (adjusted for punctuation and repetition removal)
 def transcribe_audio(audio_path):
     waveform, sample_rate = torchaudio.load(audio_path)
 
-    if waveform.shape[0] > 1:
-        waveform = torch.mean(waveform, dim=0, keepdim=True)
+    # Convert stereo to mono (if needed)
+    if waveform.shape[0] > 1:  # More than 1 channel
+        waveform = torch.mean(waveform, dim=0, keepdim=True)  # Average the channels
 
-    waveform = waveform.squeeze(0).numpy()
+    waveform = waveform.squeeze(0).numpy()  # Convert to NumPy (1D array)
 
     duration = waveform.shape[0] / sample_rate
     if duration > 60:
-        chunk_size = sample_rate * 55
-        step_size = sample_rate * 50
+        chunk_size = sample_rate * 55  # 55 seconds
+        step_size = sample_rate * 50   # 50 seconds overlap
         results = []
 
         for start in range(0, waveform.shape[0], step_size):