Update app.py

app.py

@@ -1,17 +1,12 @@
-# Imports
-import spaces
 import os
 import streamlit as st
 import requests
 from transformers import pipeline
-import openai
-
-# Suppressing all warnings
-import warnings
-warnings.filterwarnings("ignore")
+from together import Together
+from typing import Dict
 
 # Image-to-text
-def img2txt(url):
+def img2txt(url: str) -> str:
     print("Initializing captioning model...")
     captioning_model = pipeline("image-to-text", model="Salesforce/blip-image-captioning-base")
     
@@ -22,42 +17,54 @@ def img2txt(url):
     return text
 
 # Text-to-story
-def txt2story(img_text, top_k, top_p, temperature):
-
-    headers = {"Authorization": f"Bearer {os.environ['TOGETHER_API_KEY']}"}
-
-    data = {
-        "model": "togethercomputer/llama-2-70b-chat",
-        "messages": [
-            {"role": "system", "content": '''As an experienced short story writer, write story title and then create a meaningful story influenced by provided words. 
-        Ensure stories conclude positively within 100 words. Remember the story must end within 100 words''', "temperature": temperature},
-            {"role": "user", "content": f"Here is input set of words: {img_text}", "temperature": temperature} 
+def txt2story(prompt: str, top_k: int, top_p: float, temperature: float) -> str:
+    client = Together(api_key=os.environ.get("TOGETHER_API_KEY"))
+    stream = client.chat.completions.create(
+        model="meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo",
+        messages=[
+            {"role": "system", "content": '''As an experienced short story writer, write a story title and then create a meaningful story influenced by the provided prompt.
+            Ensure the story is full of positive inspiration & enthusiasm and concludes with a happy ending within 250 words. Remember the story must end within 100 words'''},
+            {"role": "user", "content": prompt}
         ],
-        "top_k": top_k,
-        "top_p": top_p,
-        "temperature": temperature
-    }
+        top_k=top_k,
+        top_p=top_p,
+        temperature=temperature,
+        stream=True
+    )
     
-    response = requests.post("https://api.together.xyz/inference", headers=headers, json=data)
+    story = ''
+    for chunk in stream:
+        story += chunk.choices[0].delta.content
     
-    story = response.json()["output"]["choices"][0]["text"] 
     return story
 
-
 # Text-to-speech
-def txt2speech(text):
+def txt2speech(text: str) -> None:
     print("Initializing text-to-speech conversion...")
     API_URL = "https://api-inference.huggingface.co/models/espnet/kan-bayashi_ljspeech_vits"
     headers = {"Authorization": f"Bearer {os.environ['HUGGINGFACEHUB_API_TOKEN']}"}
     payloads = {'inputs': text}
-
+    
     response = requests.post(API_URL, headers=headers, json=payloads)
     
     with open('audio_story.mp3', 'wb') as file:
         file.write(response.content)
 
-        
-# Streamlit web app main function
+def get_user_preferences() -> Dict[str, str]:
+    preferences = {}
+    
+    preferences['genre'] = st.selectbox("Genre", ["Science Fiction", "Fantasy", "Mystery", "Romance"])
+    preferences['setting'] = st.selectbox("Setting", ["Future", "Medieval times", "Modern day", "Alternate reality"])
+    preferences['plot'] = st.selectbox("Plot", ["Hero's journey", "Solving a mystery", "Love story", "Survival"])
+    preferences['tone'] = st.selectbox("Tone", ["Serious", "Light-hearted", "Humorous", "Dark"])
+    preferences['theme'] = st.selectbox("Theme", ["Self-discovery", "Redemption", "Love", "Justice"])
+    preferences['conflict'] = st.selectbox("Conflict Type", ["Person vs. Society", "Internal struggle", "Person vs. Nature", "Person vs. Person"])
+    preferences['magic_tech'] = st.selectbox("Magic/Technology", ["Advanced technology", "Magic system", "Supernatural abilities", "Alien technology"])
+    preferences['twist'] = st.selectbox("Mystery/Twist", ["Plot twist", "Hidden identity", "Unexpected ally/enemy", "Time paradox"])
+    preferences['ending'] = st.selectbox("Ending", ["Bittersweet", "Happy", "Open-ended", "Tragic"])
+    
+    return preferences
+
 def main():
     st.set_page_config(page_title="🎨 Image-to-Audio Story 🎧", page_icon="🖼️")
     st.title("Turn the Image into Audio Story")
@@ -71,6 +78,10 @@ def main():
     top_p = st.sidebar.number_input("Top-P", min_value=0.0, max_value=1.0, value=0.8)
     temperature = st.sidebar.number_input("Temperature", min_value=0.1, max_value=2.0, value=1.5)
 
+    # Get user preferences for the story
+    st.markdown("## Story Preferences")
+    preferences = get_user_preferences()
+
     if uploaded_file is not None:
         # Reads and saves uploaded image file
         bytes_data = uploaded_file.read()
@@ -82,7 +93,15 @@ def main():
         # Initiates AI processing and story generation
         with st.spinner("## 🤖 AI is at Work! "):
             scenario = img2txt("uploaded_image.jpg")  # Extracts text from the image
-            story = txt2story(scenario, top_k, top_p, temperature)  # Generates a story based on the image text, LLM params
+            
+            # Modify the prompt to include user preferences
+            prompt = f"Based on the image description: '{scenario}', create a {preferences['genre']} story set in {preferences['setting']}. " \
+                     f"The story should have a {preferences['tone']} tone and explore the theme of {preferences['theme']}. " \
+                     f"The main conflict should be {preferences['conflict']}. " \
+                     f"Include {preferences['magic_tech']} as a key element. " \
+                     f"The story should have a {preferences['twist']} and end with a {preferences['ending']} ending."
+            
+            story = txt2story(prompt, top_k, top_p, temperature)  # Generates a story based on the image text, LLM params, and user preferences
             txt2speech(story)  # Converts the story to audio
 
             st.markdown("---")