replaced waveform with .audio[0]

app.py

@@ -27,9 +27,9 @@ def translate(audio):
 def synthesise(text):
     inputs = tokenizer(text, return_tensors="pt")
     with torch.no_grad():
-        speech = model(inputs["input_ids"].to(device))
-    print(dir(speech))
-    return speech.waveform
+        output = model(inputs["input_ids"].to(device))
+    speech = output.audio[0]
+    return speech.cpu()
 
 
 def speech_to_speech_translation(audio):