unsqueeze

app.py

@@ -28,7 +28,7 @@ def synthesise(text):
     inputs = tokenizer(text, return_tensors="pt")
     with torch.no_grad():
         outputs = model(inputs["input_ids"])
-    speech = outputs["waveform"]
+    speech = outputs["waveform"][0]
     logging.info(speech)
     return speech.cpu()