Update app.py

app.py

@@ -91,14 +91,15 @@ def whisper_speech_demo(multilingual_text, speaker_audio):
         audio_np = generate_segment_audio(text_str, lang, speaker_url, pipe)
         print("Audio segment shape:", audio_np.shape)  # Debug statement
         audio_segments.append(audio_np)
+
     concatenated_audio = concatenate_audio_segments(audio_segments)
     print("Final concatenated audio shape:", concatenated_audio.shape)  # Debug statement
-#   audio_stereo = np.stack((concatenated_audio, concatenated_audio), axis=-1)
-#   audio_stereo = audio_stereo.reshape(-1, 2)
 
-    with tempfile.NamedTemporaryFile(suffix='.wav', delete=False) as tmp_file:
-        sf.write(tmp_file.name, concatenated_audio, 24000, format='WAV', subtype='PCM_16')
-        return tmp_file.name
+    # Normalize the concatenated audio
+    concatenated_audio = concatenated_audio / np.max(np.abs(concatenated_audio))
+
+    # Return the concatenated audio as a NumPy array
+    return concatenated_audio
 
 with gr.Blocks() as demo:
     gr.Markdown(title)