Spaces:
Sleeping
Sleeping
| # Imports | |
| import streamlit as st | |
| from transformers import pipeline | |
| from PIL import Image | |
| import torch | |
| import os | |
| import tempfile | |
| import sys | |
| import subprocess | |
| # Try to import gTTS, install if missing | |
| try: | |
| from gtts import gTTS | |
| except ImportError: | |
| st.warning("Installing required package: gTTS...") | |
| subprocess.check_call([sys.executable, "-m", "pip", "install", "gTTS"]) | |
| from gtts import gTTS | |
| st.success("gTTS installed successfully!") | |
| # Simple image-to-text function | |
| def img2text(image): | |
| image_to_text = pipeline("image-to-text", model="sooh-j/blip-image-captioning-base") | |
| text = image_to_text(image)[0]["generated_text"] | |
| return text | |
| # Improved text-to-story function with natural ending | |
| def text2story(text): | |
| generator = pipeline("text-generation", model="TinyLlama/TinyLlama-1.1B-Chat-v1.0") | |
| prompt = f"Write a short children's story based on this: {text}. The story should have a clear beginning, middle, and end. Keep it under 150 words. Once upon a time, " | |
| # Generate a longer text to ensure we get a complete story | |
| story_result = generator( | |
| prompt, | |
| max_length=300, | |
| num_return_sequences=1, | |
| temperature=0.7, | |
| do_sample=True | |
| ) | |
| story_text = story_result[0]['generated_text'] | |
| story_text = story_text.replace(prompt, "Once upon a time, ") | |
| # Find natural ending points (end of sentences) | |
| periods = [i for i, char in enumerate(story_text) if char == '.'] | |
| question_marks = [i for i, char in enumerate(story_text) if char == '?'] | |
| exclamation_marks = [i for i, char in enumerate(story_text) if char == '!'] | |
| # Combine all ending punctuation and sort | |
| all_endings = sorted(periods + question_marks + exclamation_marks) | |
| # If we have any sentence endings | |
| if all_endings: | |
| # Get the index where the story should reasonably end (after at least 100 characters) | |
| min_story_length = 100 | |
| suitable_endings = [i for i in all_endings if i >= min_story_length] | |
| if suitable_endings: | |
| # Find an ending that completes a thought (not just the first sentence) | |
| if len(suitable_endings) > 2: | |
| # Use the third sentence ending or later for a more complete story | |
| return story_text[:suitable_endings[2]+1] | |
| else: | |
| # If we don't have many sentences, use the last one we found | |
| return story_text[:suitable_endings[-1]+1] | |
| # If no good ending is found, return as is | |
| return story_text | |
| # Updated text-to-audio function using gTTS | |
| def text2audio(story_text): | |
| # Create a temporary file | |
| temp_file = tempfile.NamedTemporaryFile(delete=False, suffix='.mp3') | |
| temp_filename = temp_file.name | |
| temp_file.close() | |
| # Use gTTS to convert text to speech | |
| tts = gTTS(text=story_text, lang='en', slow=False) | |
| tts.save(temp_filename) | |
| # Read the audio file | |
| with open(temp_filename, 'rb') as audio_file: | |
| audio_bytes = audio_file.read() | |
| # Clean up the temporary file | |
| os.unlink(temp_filename) | |
| return audio_bytes | |
| # Basic Streamlit interface | |
| st.title("Image to Audio Story") | |
| uploaded_file = st.file_uploader("Upload an image") | |
| if uploaded_file is not None: | |
| # Display image | |
| st.image(uploaded_file, caption="Uploaded Image") | |
| # Convert to PIL Image | |
| image = Image.open(uploaded_file) | |
| # Image to Text | |
| with st.spinner("Generating caption..."): | |
| caption = img2text(image) | |
| st.write(f"Caption: {caption}") | |
| # Text to Story | |
| with st.spinner("Creating story..."): | |
| story = text2story(caption) | |
| st.write(f"Story: {story}") | |
| # Text to Audio | |
| with st.spinner("Generating audio..."): | |
| try: | |
| audio_bytes = text2audio(story) | |
| # Play audio | |
| st.audio(audio_bytes, format='audio/mp3') | |
| except Exception as e: | |
| st.error(f"Error generating or playing audio: {e}") | |
| st.info("If you're having issues with gTTS, you might need to manually install it with: pip install gTTS") | |
| # Fallback to a simple TTS if gTTS fails | |
| try: | |
| st.write("Attempting fallback to pyttsx3...") | |
| import pyttsx3 | |
| engine = pyttsx3.init() | |
| # Create a temporary file for the fallback audio | |
| temp_wav = tempfile.NamedTemporaryFile(delete=False, suffix='.wav') | |
| temp_wav_filename = temp_wav.name | |
| temp_wav.close() | |
| # Generate and save speech | |
| engine.save_to_file(story, temp_wav_filename) | |
| engine.runAndWait() | |
| # Read the audio file | |
| with open(temp_wav_filename, 'rb') as audio_file: | |
| fallback_audio = audio_file.read() | |
| # Clean up | |
| os.unlink(temp_wav_filename) | |
| st.audio(fallback_audio, format='audio/wav') | |
| except: | |
| st.error("Both TTS methods failed. Please install gTTS manually.") |