Merge pull request younesbram#2 from younesbram/GPT-3.5-turbo

Upgraded to gpt-3.5-turbo with few-shot training on chat completion

README.md

@@ -10,26 +10,24 @@ How to use :
 2. Attach your openai api key on client side or in env
 3. python serenitynow.py => enter any topic you'd like a seinfeld skit/joke on
 
-## TODO-asap
-yea so sometimes if the output is too long it bugs out the tortoise input. will fix
 
-I use :
+## Tech Stack
 
-OpenAI===text-davinci003
+OpenAI===gpt-3.5-turbo
 TorToiSe==2.4.2
-Conda Env
+Conda Environment
 Python / Pip
 More soon
 
 ## To collab
 please fork and follow general PR rules. 
 
 ## Intro:
-The Seinfeld-Talkabot is a project that utilizes OpenAI's davinci003 model and the TorToiSe library to generate audio files of Seinfeld characters speaking user-provided text. The goal of the project is to showcase the potential of such apps, which could be used for a variety of purposes, including generating accurate and authentic-sounding audiobooks using prompt engineering and TorToiSe, or translating and dubbing content using the GPT-3 model while maintaining the original voice. By demonstrating the capabilities of these technologies, the Seinfeld-Talkabot aims to inspire further innovation in the field of natural language processing and audio generation.
+The Seinfeld-Talkabot is a project that utilizes OpenAI's gpt-3.5-turbo model and the TorToiSe library to generate audio files of Seinfeld characters speaking user-provided text. The goal of the project is to showcase the potential of such apps, which could be used for a variety of purposes, including generating accurate and authentic-sounding audiobooks using prompt engineering and TorToiSe, or translating and dubbing content using the GPT-3 model while maintaining the original voice. By demonstrating the capabilities of these technologies, the Seinfeld-Talkabot aims to inspire further innovation in the field of natural language processing and audio generation.
 
 
 ## Todo:
-Create any kind of frontend to interact with backend of selecting a topic (Text input) => davinci003 generate a script => api call to TorToiSe + pyfastmp3encoder to generate a .wav file with the selected seinfeld character voice. Currently supporting jerry seinfelds voice from his audiobook + wav files from the seinfeld show itself.
+Create any kind of frontend to interact with backend of selecting a topic (Text input) => turbogpt3.5 generate a script => api call to TorToiSe + pyfastmp3encoder to generate a .wav file with the selected seinfeld character voice. Currently supporting jerry seinfelds voice from his audiobook + wav files from the seinfeld show itself.
 
 ## Ethical Promises
 This project is just to showcase the potential of such apps. Hopefully in the future we will have a more clear guideline on ethically using applications such as this. I hope I gave enough credit where due. Please let me know of any issues [email protected]

serenitynow.py

@@ -3,23 +3,29 @@
 import subprocess
 import uuid
 
-openai.api_key = YOUR_OPENAI_KEY_API
+openai.api_key = YOUR_API_KEY
 # Get user input for the topic
 topic = input("Enter a topic: ")
 
-# Generate text using OpenAI's GPT-3 API
-prompt = f"Generate a topic about {topic} described by a sarcastic Jerry Seinfeld and give the text."
-response = openai.Completion.create(
-    engine="text-davinci-003",
-    prompt=prompt,
-    max_tokens=2048,
-    n=1,
-    stop=None,
-    temperature=0.6,
+# A faked few-shot conversation to prime the model into becoming a sarcastic seinfeld
+response = openai.ChatCompletion.create(
+    model="gpt-3.5-turbo",
+    messages=[
+        {"role": "system", "content": "You are an assistant tasked with making Jerry Seinfeld jokes. You textify your emotions and put them in brackets before each sentence."},
+        {"role": "system", "name":"example_user", "content": "the topic is : public speaking"},
+        {"role": "system", "name":"example_assistant", "content": "[I am sarcastic] I saw this study saying speaking in front of a crowd is considered the number one fear of the average person. I found that amazing. [I am sarcastic and surprised] Number two.. was death. [I am surprised and sarcastic] Death is number two?! [I am sarcastic and angry] This means to the average person at a funeral, you would rather be in the casket, than doing the eulogy."},
+        {"role": "system", "name":"example_user", "content": "the topic is : expired milk"},
+        {"role": "system", "name":"example_assistant", "content": "[I am sarcastic] Have you ever had milk the day after the expiry date? Scares the hell out of you, doesn't it. [I am scared and sarcastic] It's after the day! I'm taking a really big chance here! [I am sarcastic and excited] Makes you wonder, how are the dates exact? Maybe the cows tip them off when they're milking them."},
+        {"role": "system", "name":"example_user", "content": "That was perfect! Great job jerry."},
+        {"role": "system", "name":"example_assistant", "content": "Thank you! It was a pleasure to deliver a sarcastic take on the topic."},
+        {"role": "user", "content": f"the topic is : {topic}."},
+    ],
+    temperature=0.666,
 )
 
+
 # Get the generated text from the response
-generated_text = response.choices[0].text.strip()
+generated_text = response['choices'][0]['message']['content']
 print(generated_text)