Update app.py

app.py

@@ -1,12 +1,54 @@
 import openai
 import gradio as gr
+import time
 import warnings
+import warnings
+import os
+from gtts import gTTS
+from transformers import AutoModelForCausalLM, AutoTokenizer
+import torch
 
 warnings.filterwarnings("ignore")
+
 openai.api_key = "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"
 
 
+def chatgpt_api(input_text):
+    '''messages = [
+    {"role": "system", "content": "You are a helpful assistant."}]
+    
+    if input_text:
+        messages.append(
+            {"role": "user", "content": input_text},
+        )
+        chat_completion = openai.ChatCompletion.create(
+            model="gpt-3.5-turbo", messages=messages
+        )
+    
+    reply = chat_completion.choices[0].message.content'''
+    input_ids = tokenizer.encode(text + tokenizer.eos_token, return_tensors="pt")
+    # concatenate new user input with chat history (if there is)
+    bot_input_ids = torch.cat([chat_history_ids, input_ids], dim=-1)
+    # generate a bot response
+    chat_history_ids = model.generate(
+        bot_input_ids,
+        max_length=1000,
+        do_sample=True,
+        top_p=0.95,
+        top_k=0,
+        temperature=0.75,
+        pad_token_id=tokenizer.eos_token_id
+    )
+    #print the output
+    reply = tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)
+
+    return reply
+
+#ffmpeg -f lavfi -i anullsrc=r=44100:cl=mono -t 10 -q:a 9 -acodec libmp3lame Temp.mp3'
+
 def transcribe(audio, text):
+    language = "en"
+
     if audio is not None:
         with open(audio, "rb") as transcript:
             prompt = openai.Audio.transcribe("whisper-1", transcript)
@@ -22,18 +64,25 @@ def transcribe(audio, text):
         stop=None,
         temperature=0.5,
     )
-    r = response.choices[0].text
+    
+    out_result = chatgpt_api(s)
+    
+    audioobj = gTTS(text = out_result, 
+                    lang = language,  
+                    slow = False)
+    
+    audioobj.save("Temp.mp3")
 
-    return [s,r]
+    return [s, out_result, "Temp.mp3"]
 
 
 with gr.Blocks() as demo:
-    gr.Markdown("Boba Jarvis strikes again")
-    with gr.Row():
-        input1 = gr.inputs.Audio(source="microphone", type = "filepath", label="Use your voice to chat")
-        input2 = gr.inputs.Textbox(lines=7, label="Chat with AI")
-        output_1 = gr.Textbox(label = "Speech to Text")
-        output_2 = gr.Textbox(label = "AI Output")
+    gr.Markdown("Dilip can finally talk!?")
+    input1 = gr.inputs.Audio(source="microphone", type = "filepath", label="Use your voice to chat")
+    input2 = gr.inputs.Textbox(lines=7, label="Chat with AI")
+    output_1 = gr.Textbox(label="Text Input")
+    output_2 = gr.Textbox(label="Text Output")
+    output_3 = gr.Audio("Temp.mp3", label="Speech Output")
     btn = gr.Button("Run")
-    btn.click(fn=transcribe, inputs=[input1, input2], outputs=[output_1, output_2])
-demo.launch() 
+    btn.click(fn=transcribe, inputs=[input1, input2], outputs=[output_1, output_2, output_3])
+demo.launch(share=True)