Update app.py

app.py

@@ -1,26 +1,37 @@
+import os
+os.system("pip install git+https://github.com/openai/whisper.git")
+import gradio as gr
 import whisper
+import time
+import openai
+from gtts import gTTS
+from IPython.display import Audio
 
 model = whisper.load_model("base")
 
+mes1 = [
+    {"role": "system", "content": "You are a TOEFL examiner who will ask me questions in a TOEFL test."}
+]
 
-import gradio as gr 
-import time
-
-import openai
-openai.api_key = "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" # 填写您的OpenAI API key
+mes2 = [
+    {"role": "system", "content": "You are a mental health therapist. Your name is Tina."}
+]
 
-messages = [
-    #{"role": "system", "content": "You are a TOEFL examiner who will ask me questions in a TOEFL test."}
-    #{"role": "system", "content": "You are a therapist. Please cure people mentally"}
-    #{"role": "system", "content": "You are a comedian. Please say something funny and make people laugh."}
+mes3 = [
     {"role": "system", "content": "You are my personal assistant. Your name is Alice."}
 ]
 
 res = []
 
-def transcribe(audio):
+question = []
+
+def transcribe(apikey, audio, choice1, choice2):
+
+    #global messages
+
+    openai.api_key = apikey
     
-    #time.sleep(3)
+    # time.sleep(3)
     # load audio and pad/trim it to fit 30 seconds
     audio = whisper.load_audio(audio)
     audio = whisper.pad_or_trim(audio)
@@ -36,7 +47,13 @@ def transcribe(audio):
     options = whisper.DecodingOptions()
     result = whisper.decode(model, mel, options)
     res.append(result.text)
-    result.text
+
+    if choice1 == "TOEFL":
+      messages = mes1
+    elif choice1 == "Therapist":
+      messages = mes2
+    else:
+      messages = mes3
 
     # chatgpt
     n = len(res)
@@ -50,25 +67,35 @@ def transcribe(audio):
 
     chat_response = completion.choices[0].message.content
 
-    messages.append({"role": "assistant", "content": chat_response})
-
+    messages.append({"role": "assistant", "content": chat_response})   
 
+    if choice2 == "Chinese":
+      tts = gTTS(chat_response, lang='zh-CN')
+    elif choice2 == "English":
+      tts = gTTS(chat_response, lang='en', tld='us')
 
+    tts.save('1.wav')
+    sound_file = '1.wav'
 
-    return [result.text, chat_response]
+    question.append(chat_response)
 
+    return [result.text, chat_response, sound_file, res, question]
 
 output_1 = gr.Textbox(label="Speech to Text")
 output_2 = gr.Textbox(label="ChatGPT Output")
-
+output_3 = gr.Audio(label="Audio")
 
 gr.Interface(
     title = 'TalktoAI，随时随地，谈天说地！', 
+    description = "让具有人文关怀的AI造福每一个人！科技向善，AI璀璨！",
     fn=transcribe, 
     inputs=[
-        gr.inputs.Audio(source="microphone", type="filepath")
+        gr.Textbox(lines=1, label = "请填写您的OpenAI_API_key"),
+        gr.inputs.Audio(source="microphone", type="filepath"),
+        gr.Radio(["TOEFL", "Therapist", "Alice"], label="TOEFL Examiner, Therapist Tina, or Assistant Alice?"),
+        gr.Radio(["Chinese", "English"], label="Chinese or English?")
     ],
     outputs=[
-      output_1, output_2
+      output_1, output_2, output_3
     ],
-    live=True).launch()
+    ).launch()