Update app.py

app.py

@@ -5,7 +5,8 @@ from datasets import load_dataset
 import librosa
 from transformers import SpeechT5ForTextToSpeech, SpeechT5HifiGan, SpeechT5Processor, pipeline
 
-
+target_dtype = np.int16
+max_range = np.iinfo(target_dtype).max
 device = "cuda:0" if torch.cuda.is_available() else "cpu"
 
 # load speech translation checkpoint
@@ -40,8 +41,8 @@ def speech_to_speech_translation(audio):
     data_16 = librosa.resample(data_array, orig_sr=samplerate, target_sr=sampling_rate)
     translated_text = translate(data_16)
     synthesised_speech = synthesise(translated_text)
-    synthesised_speech = (synthesised_speech.numpy() * 32767).astype(np.int16)
-    return 16000, synthesised_speech
+    synthesised_speech = (synthesised_speech.numpy() * max_range).astype(np.int16)
+    return sampling_rate, synthesised_speech
 
 
 title = "Cascaded STST"