Update app.py

app.py

@@ -21,6 +21,7 @@ from PIL import Image
 import io
 from gliner import GLiNER
 import tempfile
+from pydub import AudioSegment
 
 nltk.download("punkt")
 nltk.download("punkt_tab")
@@ -219,13 +220,23 @@ def generate_audio_kokoro(text, lang, selected_voice):
     audio_data_list = [audio for _, _, audio in generator]
     full_audio = np.concatenate(audio_data_list)
     
-	# Save to a temporary file
-    with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as temp_file:
-        sf.write(temp_file, full_audio, 24000, format='wav')
-        temp_file_path = temp_file.name
+	# Save as WAV (Temporary Step for Conversion)
+    temp_wav_file = f"temp_audio_{lang}.wav"
+    sf.write(temp_wav_file, full_audio, 24000)
+
+    # Convert WAV to MP3
+    output_file = f"audio_{lang}.mp3"
+    AudioSegment.from_wav(temp_wav_file).export(output_file, format="mp3", bitrate="192k")
+
+    # Clean up the temporary WAV file
+    os.remove(temp_wav_file)
 	
-    print("Audio generated successfully.")
-    return temp_file_path
+	# Confirm file is saved successfully
+    if not os.path.exists(output_file):
+        raise FileNotFoundError(f"Audio file '{output_file}' was not created successfully.")
+
+    # Return MP3 file path for Gradio Audio Box (with timestamp for cache bypass)
+    return f"{output_file}?{int(time.time())}"
 	
 ### 5️⃣ Chunk-Based Summarization
 def split_text_with_optimized_overlap(text, max_tokens=1024, overlap_tokens=25):