Update app.py

app.py

@@ -40,7 +40,8 @@ def process_audio(audio_file):
     
         try:
             audio_segment = AudioSegment.from_mp3(audio_file)
-            wav_path = audio_file.replace(".mp3", ".wav")
+            wav_path = "/tmp/" + audio_file.split("/")[-1].replace(".mp3", ".wav")  # 將檔案存放於 /tmp 目錄
+            
             audio_segment.export(wav_path, format="wav")
             
         except Exception as e:
@@ -48,37 +49,45 @@ def process_audio(audio_file):
         
         # 語音轉文字
             
-        result= whisper_pipe(wav_path)["text"]
+        try:
+                result= whisper_pipe(wav_path)["text"]
                 
-        # 使用 T5 作為替代模型
-        
-        messages=[{"role": "user", "content": result}]
-        
-        deepseek_response="" 
-        
-        try: 
-            from transformers import pipeline
-            
-            pipe=pipeline("text-generation",model="t5-base")
-            
-            deepseek_response=pipe(messages)[0]["generated_text"]
-                    
-                # 使用 spaCy 分析文本
+                # 使用 T5 作為替代模型
+                
+                messages=[{"role": "user", "content": result}]
+                
+                deepseek_response="" 
+                
+                try: 
+                    from transformers import pipeline
                     
-            doc=nlp(deepseek_response) if nlp is not None else None 
-            entities=[(ent.text, ent.label_) for ent in doc.ents] if doc is not None else []
+                    pipe=pipeline("text-generation",model="t5-base")
                     
-            return {
-                "Transcription (Whister)": result,
-                "AI Response (T5)": deepseek_response,# 修改為 T5 回應以避免與原來不同步                            
-                "Extracted Entities (spaCy)": entities}
+                    deepseek_response=pipe(messages)[0]["generated_text"]
+                            
+                        # 使用 spaCy 分析文本
                         
-        except Exception as e: 
+                    doc=nlp(deepseek_response) if nlp is not None else None 
+                    entities=[(ent.text, ent.label_) for ent in doc.ents] if doc is not None else []
+                            
                     return {
-                        "Transcription (Whister)": result,# 保留原始轉錄內容                                
-                            }
+                        "Transcription (Whister)": result,
+                        "AI Response (T5)": deepseek_response,# 修改為 T5 回應以避免與原來不同步                            
+                        "Extracted Entities (spaCy)": entities}
                             
+                except Exception as e: 
+                            return {
+                                "Transcription (Whister)": result,# 保留原始轉錄內容                                
+                                    }
+                                
+        
+        except Exception as e:
+            
+            return {"Error": f"語音轉文字失敗：{e}"}
 
+def clear_input():
+    
+        return "", ""
 
 with gr.Blocks() as app:
 
@@ -96,10 +105,4 @@ with gr.Blocks() as app:
   
   
 
-   submit_button = gr.Button("提交")  
-   
-   submit_button.click(fn=lambda x: process_audio(x), inputs=[audio_input], outputs=[output_text])  
-
-
-if __name__ == "__main__":
-     app.launch()
+   submit