added sentencepiece tokenizer and updated requirements.txt

app.py

@@ -50,15 +50,30 @@ def synthesise_audio(text, forward_params=None):
     #    audio = f.getvalue()
 
     # Convert numpy array to audio
-    audio = np.int16(speech["audio"] * 32767)
-    audio_segment = AudioSegment(audio, sample_width=2, frame_rate=speech["sampling_rate"], channels=1)
-
+    #audio = np.int16(speech["audio"] * 32767)
+    #audio_segment = AudioSegment(audio, sample_width=2, frame_rate=speech["sampling_rate"], channels=1)
 
+    # Convert numpy array to list
+    #audio = speech["audio"]
 
     #return speech["audio"]
 
-    return audio
-    
+    #return audio
+
+
+    # Create an in-memory buffer to store the audio data
+    audio_buffer = io.BytesIO()
+
+    # Write the audio data to the in-memory buffer
+    sf.write(audio_buffer, speech["audio"], samplerate=speech["sampling_rate"], format="WAV")
+
+    # Move the buffer cursor to the beginning of the buffer
+    audio_buffer.seek(0)
+
+    # Read the audio data from the in-memory buffer into a numpy array
+    audio, sr = sf.read(audio_buffer)
+
+    return audio, sr    
   
     
 

requirements.txt

@@ -1,3 +1,4 @@
 transformers
 soundfile
 torch
+sentencepiece