Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -3,15 +3,8 @@ import streamlit as st
|
|
3 |
from transformers import pipeline
|
4 |
from gtts import gTTS
|
5 |
import os
|
6 |
-
from retrying import retry # For retry logic
|
7 |
-
|
8 |
-
# Retry decorator for handling transient errors
|
9 |
-
def retry_if_timeout_error(exception):
|
10 |
-
"""Return True if the exception is a timeout error."""
|
11 |
-
return isinstance(exception, Exception) and "Read timed out" in str(exception)
|
12 |
|
13 |
# Function to convert image to text using Hugging Face's BLIP model
|
14 |
-
@retry(stop_max_attempt_number=3, wait_fixed=2000, retry_on_exception=retry_if_timeout_error) # Retry 3 times with a 2-second delay
|
15 |
def img2text(url):
|
16 |
"""
|
17 |
Converts an image to text using the Salesforce/blip-image-captioning-base model.
|
@@ -22,8 +15,7 @@ def img2text(url):
|
|
22 |
Returns:
|
23 |
str: Generated text caption from the image, without words like "illustration".
|
24 |
"""
|
25 |
-
|
26 |
-
image_to_text_model = pipeline("image-to-text", model="Salesforce/blip-image-captioning-base", device_map="auto", timeout=30)
|
27 |
text = image_to_text_model(url)[0]["generated_text"]
|
28 |
|
29 |
# Remove unwanted words like "illustration"
|
@@ -34,7 +26,6 @@ def img2text(url):
|
|
34 |
return text.strip()
|
35 |
|
36 |
# Function to generate a kid-friendly story from the text caption
|
37 |
-
@retry(stop_max_attempt_number=3, wait_fixed=2000, retry_on_exception=retry_if_timeout_error) # Retry 3 times with a 2-second delay
|
38 |
def text2story(text):
|
39 |
"""
|
40 |
Generates a kid-friendly story from the text caption using the aspis/gpt2-genre-story-generation model.
|
@@ -48,8 +39,8 @@ def text2story(text):
|
|
48 |
# Add a prompt to ensure the story is happy, fun, and appropriate for kids
|
49 |
prompt = f"Base on the following scenario: {text}, write a happy,fun and complete story for kids with no sad, violent, scary elements."
|
50 |
|
51 |
-
# Load the text generation model
|
52 |
-
story_generator = pipeline("text-generation", model="aspis/gpt2-genre-story-generation"
|
53 |
|
54 |
# Generate the story
|
55 |
story = story_generator(prompt, max_length=150, num_return_sequences=1, temperature=0.7, top_k=50, top_p=0.9)[0]["generated_text"]
|
@@ -57,7 +48,7 @@ def text2story(text):
|
|
57 |
# Remove the prompt from the generated story
|
58 |
story = story.replace(prompt, "").strip()
|
59 |
|
60 |
-
# Ensure the story is within
|
61 |
story_words = story.split()
|
62 |
if len(story_words) > 100:
|
63 |
story = " ".join(story_words[:100])
|
@@ -110,31 +101,21 @@ if uploaded_file is not None:
|
|
110 |
|
111 |
# Stage 1: Image to Text
|
112 |
st.text('β¨ Turning your picture into words...')
|
113 |
-
|
114 |
-
|
115 |
-
st.write("**What we see:**", scenario)
|
116 |
-
except Exception as e:
|
117 |
-
st.error(f"Oops! Something went wrong while processing the image. Please try again. Error: {e}")
|
118 |
|
119 |
# Stage 2: Text to Story
|
120 |
st.text('π Creating a fun story for you...')
|
121 |
-
|
122 |
-
|
123 |
-
st.write("**Your story:**", story)
|
124 |
-
except Exception as e:
|
125 |
-
st.error(f"Oops! Something went wrong while generating the story. Please try again. Error: {e}")
|
126 |
|
127 |
# Stage 3: Story to Audio
|
128 |
st.text('π§ Turning your story into audio...')
|
129 |
-
|
130 |
-
audio_file = text2audio(story)
|
131 |
-
except Exception as e:
|
132 |
-
st.error(f"Oops! Something went wrong while converting the story to audio. Please try again. Error: {e}")
|
133 |
|
134 |
# Play button for the generated audio
|
135 |
if st.button("π΅ **Play Audio**"):
|
136 |
st.audio(audio_file, format="audio/mp3")
|
137 |
|
138 |
# Clean up the generated audio file
|
139 |
-
|
140 |
-
os.remove(audio_file)
|
|
|
3 |
from transformers import pipeline
|
4 |
from gtts import gTTS
|
5 |
import os
|
|
|
|
|
|
|
|
|
|
|
|
|
6 |
|
7 |
# Function to convert image to text using Hugging Face's BLIP model
|
|
|
8 |
def img2text(url):
|
9 |
"""
|
10 |
Converts an image to text using the Salesforce/blip-image-captioning-base model.
|
|
|
15 |
Returns:
|
16 |
str: Generated text caption from the image, without words like "illustration".
|
17 |
"""
|
18 |
+
image_to_text_model = pipeline("image-to-text", model="Salesforce/blip-image-captioning-base")
|
|
|
19 |
text = image_to_text_model(url)[0]["generated_text"]
|
20 |
|
21 |
# Remove unwanted words like "illustration"
|
|
|
26 |
return text.strip()
|
27 |
|
28 |
# Function to generate a kid-friendly story from the text caption
|
|
|
29 |
def text2story(text):
|
30 |
"""
|
31 |
Generates a kid-friendly story from the text caption using the aspis/gpt2-genre-story-generation model.
|
|
|
39 |
# Add a prompt to ensure the story is happy, fun, and appropriate for kids
|
40 |
prompt = f"Base on the following scenario: {text}, write a happy,fun and complete story for kids with no sad, violent, scary elements."
|
41 |
|
42 |
+
# Load the text generation model
|
43 |
+
story_generator = pipeline("text-generation", model="aspis/gpt2-genre-story-generation")
|
44 |
|
45 |
# Generate the story
|
46 |
story = story_generator(prompt, max_length=150, num_return_sequences=1, temperature=0.7, top_k=50, top_p=0.9)[0]["generated_text"]
|
|
|
48 |
# Remove the prompt from the generated story
|
49 |
story = story.replace(prompt, "").strip()
|
50 |
|
51 |
+
# Ensure the story is within 95 words
|
52 |
story_words = story.split()
|
53 |
if len(story_words) > 100:
|
54 |
story = " ".join(story_words[:100])
|
|
|
101 |
|
102 |
# Stage 1: Image to Text
|
103 |
st.text('β¨ Turning your picture into words...')
|
104 |
+
scenario = img2text(uploaded_file.name)
|
105 |
+
st.write("**What we see:**", scenario)
|
|
|
|
|
|
|
106 |
|
107 |
# Stage 2: Text to Story
|
108 |
st.text('π Creating a fun story for you...')
|
109 |
+
story = text2story(scenario)
|
110 |
+
st.write("**Your story:**", story)
|
|
|
|
|
|
|
111 |
|
112 |
# Stage 3: Story to Audio
|
113 |
st.text('π§ Turning your story into audio...')
|
114 |
+
audio_file = text2audio(story)
|
|
|
|
|
|
|
115 |
|
116 |
# Play button for the generated audio
|
117 |
if st.button("π΅ **Play Audio**"):
|
118 |
st.audio(audio_file, format="audio/mp3")
|
119 |
|
120 |
# Clean up the generated audio file
|
121 |
+
os.remove(audio_file)
|
|