Update app.py

app.py

@@ -9,13 +9,11 @@ embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validat
 speaker_embeddings = embeddings_dataset[7306]["xvector"]
 speaker_embeddings = torch.tensor(speaker_embeddings).unsqueeze(0)
 
-
 # 加载 Visual Question Answering 模型 microsoft/git-base-vqav2
 vqa_pipeline = pipeline("text2text-generation", model="microsoft/git-base-vqav2")
 
 # 加载文本到语音模型并手动指定
 text_to_speech_pipeline = pipeline("text-to-speech", model="microsoft/speecht5_tts")
-audio_data = text_to_speech_pipeline(answer, speaker_embeddings=speaker_embeddings)
 
 def main():
     st.title("Visual Question Answering with Text-to-Speech")
@@ -25,7 +23,7 @@ def main():
 
     if st.button("Get Answer"):
         answer = vqa_pipeline(question, image_path)[0]['generated_text']
-        audio_data = text_to_speech_pipeline(answer)
+        audio_data = text_to_speech_pipeline(answer, speaker_embeddings=speaker_embeddings)
 
         st.write("Answer:", answer)
         st.audio(audio_data[0]["audio"], format='audio/wav')