Update app.py

app.py

@@ -12,7 +12,7 @@ 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
+# 加载文本到语音模型
 text_to_speech_pipeline = pipeline("text-to-speech", model="microsoft/speecht5_tts")
 
 def main():
@@ -23,7 +23,10 @@ def main():
 
     if st.button("Get Answer"):
         answer = vqa_pipeline(question, image_path)[0]['generated_text']
-        audio_data = text_to_speech_pipeline(answer, speaker_embeddings=speaker_embeddings)
+
+        # 将说话者的嵌入向量作为文本的一部分传递给文本到语音模型
+        text_with_speaker = f"{answer} Speaker Embeddings: {speaker_embeddings}"
+        audio_data = text_to_speech_pipeline(text_with_speaker)
 
         st.write("Answer:", answer)
         st.audio(audio_data[0]["audio"], format='audio/wav')