Assignment1 / app.py
CR7CAD's picture
Update app.py
a084b90 verified
raw
history blame
3.02 kB
# import part
import streamlit as st
from transformers import pipeline
import os
# function part
# img2text
def img2text(image_path):
image_to_text = pipeline("image-to-text", model="dumperize/movie-picture-captioning")
text = image_to_text(image_path)[0]["generated_text"]
return text
# text2story
def text2story(text):
messages = [
{"role": "user", "content": "Who are you?"},
]
# Using a smaller text generation model
generator = pipeline("text-generation", model="mlx-community/Llama-3.2-1B-Instruct-4bit")
generator(messages)
# Create a prompt for the story generation
prompt = f"Write a fun children's story based on this: {text}. Once upon a time, "
# Generate the story
story_result = generator(
prompt,
max_length=200,
num_return_sequences=1,
temperature=0.8,
top_k=50,
top_p=0.95,
do_sample=True
)
# Extract the generated text
story_text = story_result[0]['generated_text']
story_text = story_text.replace(prompt, "Once upon a time, ")
# Make sure the story is at least 100 words
words = story_text.split()
if len(words) > 100:
# Simply truncate to 100 words
story_text = " ".join(words[:100])
return story_text
# text2audio
def text2audio(story_text):
tts = pipeline("text-to-speech", model="espnet/kan-bayashi_ljspeech_vits")
audio_output = tts(story_text)
return {
"audio": audio_output["audio"],
"sampling_rate": audio_output["sampling_rate"]
}
# Function to save temporary image file
def save_uploaded_image(uploaded_file):
if not os.path.exists("temp"):
os.makedirs("temp")
image_path = os.path.join("temp", uploaded_file.name)
with open(image_path, "wb") as f:
f.write(uploaded_file.getvalue())
return image_path
# main part
st.set_page_config(page_title="Your Image to Audio Story", page_icon="🦜")
st.header("Turn Your Image to Audio Story")
uploaded_file = st.file_uploader("Select an Image...")
if uploaded_file is not None:
# Display the uploaded image
st.image(uploaded_file, caption="Uploaded Image", use_container_width=True)
# Save the image temporarily
image_path = save_uploaded_image(uploaded_file)
# Stage 1: Image to Text
st.text('Processing img2text...')
caption = img2text(image_path)
st.write(caption)
# Stage 2: Text to Story
st.text('Generating a story...')
story = text2story(caption)
st.write(story)
# Stage 3: Story to Audio data
st.text('Generating audio data...')
audio_data = text2audio(story)
# Play button
if st.button("Play Audio"):
st.audio(
audio_data["audio"],
format="audio/wav",
start_time=0,
sample_rate=audio_data["sampling_rate"]
)
# Clean up the temporary file
try:
os.remove(image_path)
except:
pass