tonyhui2234's picture
Update app.py
9cda92f verified
import streamlit as st
from transformers import pipeline
import numpy as np
import torch
from scipy.io.wavfile import write
# function part
# img2text
def img2text(url):
image_to_text_model = pipeline("image-to-text", model="Salesforce/blip-image-captioning-base")
text = image_to_text_model(url)[0]["generated_text"]
return text
# text2story
def text2story(text):
text_generation_model = pipeline("text-generation", model="openai-community/gpt2")
generator = pipeline("text-generation", model="openai-community/gpt2")
story_text = text_generation_model(text, max_length=60, truncation=True)
story = story_text[0]["generated_text"]
return story
# # text2audio
def text2audio(story_text, output_path="kids_playing_audio.wav"):
text_to_speech_model = pipeline("text-to-speech", model="facebook/mms-tts-eng", device=0) # Use GPU if available
audio_data = text_to_speech_model(story_text)
## Save the audio to kids_playing_audio.wav, so that it can play
# Extract audio waveform and sample rate
waveform = np.array(audio_data["audio"], dtype=np.float32) # Ensure correct format
sample_rate = int(audio_data["sampling_rate"]) # Convert sample rate to integer
# 🔥 Ensure sample rate is within a valid range (between 1 and 65535 Hz)
if sample_rate <= 0 or sample_rate > 96000: # 96000 is a high-res audio limit
print(f"⚠️ Warning: Invalid sample rate detected ({sample_rate} Hz). Setting to 44100 Hz.")
sample_rate = 44100 # Use standard sample rate
# Ensure waveform is a 1D array (some models return 2D arrays)
if waveform.ndim > 1:
waveform = waveform.mean(axis=0) # Convert to mono by averaging channels
# Convert float waveform (-1.0 to 1.0) to 16-bit PCM format
waveform_int16 = np.int16(waveform * 32767)
# Save the audio file
write(output_path, sample_rate, waveform_int16)
print(f"✅ Audio saved as {output_path} with sample rate {sample_rate}")
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:
print(uploaded_file)
bytes_data = uploaded_file.getvalue()
with open(uploaded_file.name, "wb") as file:
file.write(bytes_data)
st.image(uploaded_file, caption="Uploaded Image",
use_column_width=True)
#Stage 1: Image to Text
st.text('Processing img2text...')
scenario = img2text(uploaded_file.name)
st.write(scenario)
#Stage 2: Text to Story
st.text('Generating a story...')
story = text2story(scenario)
st.write(story)
#Stage 3: Story to Audio data
st.text('Generating audio data...')
text2audio(story)
# # Play button
if st.button("Play Audio"):
st.audio("kids_playing_audio.wav")