demo

app.py

@@ -44,8 +44,8 @@ def predict(text, speaker, audio):
         speaker_embedding = create_speaker_embedding(audio)
     else:
         speaker_embedding = np.load(speaker_embeddings[speaker[:3]])
-        speaker_embedding = torch.tensor(speaker_embedding).unsqueeze(0)
 
+    speaker_embedding = torch.tensor(speaker_embedding).unsqueeze(0)
     inputs = processor(text=text, return_tensors="pt")
     speech = model.generate_speech(inputs["input_ids"], speaker_embedding, vocoder=vocoder)
     speech = (speech.numpy() * 32767).astype(np.int16)