# Import necessary libraries import streamlit as st # For building the web application from transformers import pipeline # For using pre-trained models (image-to-text and text-generation) from gtts import gTTS # For converting text to speech import os # For file handling (saving and deleting temporary files) # Function to convert image to text using Hugging Face's BLIP model def img2text(url): """ Converts an image to text using the Salesforce/blip-image-captioning-base model. Args: url (str): Path to the image file. Returns: str: Generated text caption from the image, without unwanted words like "illustration". """ try: # Load the image-to-text model image_to_text_model = pipeline("image-to-text", model="Salesforce/blip-image-captioning-base") # Generate text caption from the image text = image_to_text_model(url)[0]["generated_text"] # Remove unwanted words like "illustration" to make the caption cleaner unwanted_words = ["illustration", "drawing", "sketch", "picture", "dream", "imagination"] for word in unwanted_words: text = text.replace(word, "") return text.strip() # Return the cleaned caption except Exception as e: st.error(f"Error processing image: {e}") # Display error message if something goes wrong return None # Function to generate a kid-friendly superhero story from the text caption def text2story(text): """ Generates a superhero story from the text caption using the pranavpsv/gpt2-genre-story-generator model. Args: text (str): Text caption generated from the image. Returns: str: Generated superhero story suitable for kids aged 3-10, within 100 words. """ try: # Load the text-generation model story_generator = pipeline("text-generation", model="pranavpsv/gpt2-genre-story-generator") # Generate the story with the superhero genre prompt = f" {text}" # Add genre tags to the prompt story = story_generator(prompt, max_length=150, num_return_sequences=1)[0]['generated_text'] # Remove and tags from the generated story story = story.replace("", "").replace("", "").strip() # Ensure the story is within 100 words by truncating if necessary story = " ".join(story.split()[:100]) # If the story is too short, regenerate it if len(story.split()) < 50: # Minimum 50 words story = story_generator(prompt, max_length=200, num_return_sequences=1)[0]['generated_text'] story = story.replace("", "").replace("", "").strip() story = " ".join(story.split()[:100]) return story except Exception as e: st.error(f"Error generating story: {e}") # Display error message if something goes wrong return None # Function to convert text to audio using gTTS def text2audio(story_text): """ Converts the generated story text to audio using gTTS. Args: story_text (str): The generated story text. Returns: str: Path to the generated audio file. """ try: # Convert text to speech using gTTS tts = gTTS(text=story_text, lang='en') audio_file = "story_audio.mp3" # Define the output audio file name tts.save(audio_file) # Save the audio file return audio_file except Exception as e: st.error(f"Error generating audio: {e}") # Display error message if something goes wrong return None # Main application function def main(): """ Main function to run the Streamlit application. """ # Configure the Streamlit app page st.set_page_config(page_title="Picture Stories 🎨📖", page_icon="🦄") st.title("Picture Stories 🎨📖") st.markdown("### Turn your pictures into fun superhero stories and listen to them! 🎉") # Instructions for kids st.markdown(""" **How to use this app:** 1. **Upload a picture** of something fun, like your favorite toy, your pet, or your hero. 2. Wait for the app to **create a superhero story** from your picture. 3. **Listen to the story** by clicking the "Play Audio" button. 4. Enjoy your fun superhero story! 🎧 """) # Upload image uploaded_file = st.file_uploader("📷 **Upload your picture here!**", type=["jpg", "jpeg", "png"]) # Initialize session state variables if 'scenario' not in st.session_state: st.session_state.scenario = None if 'story' not in st.session_state: st.session_state.story = None if 'audio_file' not in st.session_state: st.session_state.audio_file = None if uploaded_file is not None: # Save the uploaded file to disk image_bytes = uploaded_file.getvalue() with open(uploaded_file.name, "wb") as file: file.write(image_bytes) # Display the uploaded image in the app st.image(uploaded_file, caption="Your awesome picture!", use_container_width=True) # Stage 1: Image to Text with st.spinner('✨ Turning your picture into words...'): if st.session_state.scenario is None or uploaded_file.name != st.session_state.get('uploaded_file_name', None): st.session_state.scenario = img2text(uploaded_file.name) # Generate text caption from the image st.session_state.uploaded_file_name = uploaded_file.name # Store the uploaded file name if st.session_state.scenario: st.write("**What we see:**", st.session_state.scenario) # Display the generated caption # Stage 2: Text to Story with st.spinner('📖 Creating a fun superhero story for you...'): if st.session_state.story is None or uploaded_file.name != st.session_state.uploaded_file_name: st.session_state.story = text2story(st.session_state.scenario) # Generate a superhero story from the caption if st.session_state.story: st.write("**Your superhero story:**", st.session_state.story) # Display the generated story # Stage 3: Story to Audio with st.spinner('🎧 Turning your story into audio...'): if st.session_state.audio_file is None or uploaded_file.name != st.session_state.uploaded_file_name: st.session_state.audio_file = text2audio(st.session_state.story) # Generate audio file # Play button for the generated audio if st.button("🎵 **Play Audio**"): if os.path.exists(st.session_state.audio_file): st.audio(st.session_state.audio_file, format="audio/mp3") # Play the audio else: st.error("Audio file not found. Please try again.") # Display error if audio file is missing # Clean up temporary files (uploaded image) if os.path.exists(uploaded_file.name): os.remove(uploaded_file.name) # Delete the uploaded image file # Run the application if __name__ == "__main__": main()