T2I / app.py
VirtualOasis's picture
Update app.py
6ffa77a verified
raw
history blame
14.3 kB
import gradio as gr
import os
import tempfile
import traceback
import gc
from PIL import Image, ImageDraw, ImageFont
def draw_horizontal_lines(draw, width, height, spacing=60, color=(230, 230, 230)):
"""Draws horizontal lines on the image."""
for y in range(0, height, spacing):
draw.line([(0, y), (width, y)], fill=color, width=2)
def draw_dot_grid(draw, width, height, spacing=50, color=(220, 220, 220)):
"""Draws a dot grid on the image."""
for x in range(0, width, spacing):
for y in range(0, height, spacing):
draw.ellipse([(x-2, y-2), (x+2, y+2)], fill=color)
def draw_lattice_grid(draw, width, height, spacing=100, color=(235, 235, 235)):
"""Draws a lattice/graph paper grid on the image."""
# Draw vertical lines
for x in range(0, width, spacing):
draw.line([(x, 0), (x, height)], fill=color, width=2)
# Draw horizontal lines
for y in range(0, height, spacing):
draw.line([(0, y), (width, y)], fill=color, width=2)
def text_to_images_mcp(text_content, style='lines', font_path=None):
"""
Converts text to images and returns a list of image objects.
Args:
text_content (str): The text to be converted.
style (str): The background style ('plain', 'lines', 'dots', 'grid').
font_path (str, optional): The path to a .ttf font file.
Returns:
list: List of PIL Image objects for display in Gradio.
"""
try:
# Input validation
if not text_content or not text_content.strip():
return []
# Limit text length to prevent memory issues
if len(text_content) > 10000: # 10k character limit
text_content = text_content[:10000] + "\n\n[Text truncated due to length limit]"
# --- Configuration ---
IMG_WIDTH = 1080
IMG_HEIGHT = 1080
BACKGROUND_COLOR = (255, 255, 255)
TEXT_COLOR = (10, 10, 10)
STYLE_COLOR = (225, 225, 225)
PADDING_X = 80
PADDING_Y = 80
FONT_SIZE = 48
LINE_SPACING = 20
# Limit maximum number of pages to prevent memory issues
MAX_PAGES = 10
# --- Font Loading ---
font = None
try:
if font_path and os.path.exists(font_path):
font = ImageFont.truetype(font_path, FONT_SIZE)
else:
font_paths_to_try = [
"Arial.ttf", "arial.ttf",
"/System/Library/Fonts/Supplemental/Arial.ttf",
"/usr/share/fonts/truetype/liberation/LiberationSans-Regular.ttf",
]
for f_path in font_paths_to_try:
try:
font = ImageFont.truetype(f_path, FONT_SIZE)
break
except (IOError, OSError):
continue
if not font:
font = ImageFont.load_default()
except Exception as e:
print(f"Font loading error: {e}")
font = ImageFont.load_default()
# --- Text Wrapping Logic ---
drawable_width = IMG_WIDTH - 2 * PADDING_X
paragraphs = [p.strip() for p in text_content.strip().split('\n') if p.strip()]
if not paragraphs:
return []
all_lines_and_breaks = []
for i, paragraph in enumerate(paragraphs):
words = paragraph.split()
current_line = ""
for word in words:
try:
# Handle very long words
if hasattr(font, 'getlength') and font.getlength(word) > drawable_width:
temp_word = ""
for char in word:
if font.getlength(temp_word + char) > drawable_width:
all_lines_and_breaks.append(temp_word)
temp_word = char
else:
temp_word += char
word = temp_word
# Check line length
test_line = current_line + " " + word if current_line else word
if hasattr(font, 'getlength'):
line_width = font.getlength(test_line)
else:
line_width = len(test_line) * 12 # Rough estimate
if line_width <= drawable_width:
current_line = test_line
else:
if current_line:
all_lines_and_breaks.append(current_line.strip())
current_line = word
except Exception as e:
print(f"Text wrapping error: {e}")
# Fallback to simple character limit
if len(current_line + " " + word) <= 80: # Rough character limit
current_line += " " + word
else:
all_lines_and_breaks.append(current_line.strip())
current_line = word
if current_line:
all_lines_and_breaks.append(current_line.strip())
if i < len(paragraphs) - 1:
all_lines_and_breaks.append(None) # Paragraph break
# --- Image Generation ---
img_count = 0
page_content = []
y_text = PADDING_Y
try:
if hasattr(font, 'getbbox'):
line_height = font.getbbox("A")[3] - font.getbbox("A")[1]
else:
line_height = FONT_SIZE
except Exception:
line_height = FONT_SIZE
PARAGRAPH_SPACING = line_height
generated_images = []
def create_image_page(content, page_num):
"""Helper function to create and return a single image page."""
try:
img = Image.new('RGB', (IMG_WIDTH, IMG_HEIGHT), color=BACKGROUND_COLOR)
draw = ImageDraw.Draw(img)
# Draw the selected background style first
if style == 'lines':
line_style_spacing = line_height + LINE_SPACING
draw_horizontal_lines(draw, IMG_WIDTH, IMG_HEIGHT, spacing=line_style_spacing, color=STYLE_COLOR)
elif style == 'dots':
draw_dot_grid(draw, IMG_WIDTH, IMG_HEIGHT, color=STYLE_COLOR)
elif style == 'grid':
draw_lattice_grid(draw, IMG_WIDTH, IMG_HEIGHT, color=STYLE_COLOR)
# Draw the text on top of the background
current_y = PADDING_Y
for page_item in content:
if page_item is not None:
try:
draw.text((PADDING_X, current_y), page_item, font=font, fill=TEXT_COLOR)
current_y += line_height + LINE_SPACING
except Exception as e:
print(f"Text drawing error: {e}")
current_y += line_height + LINE_SPACING
else:
current_y += PARAGRAPH_SPACING
# Optimize image for web delivery
img = img.convert('RGB')
generated_images.append(img)
return img
except Exception as e:
print(f"Image creation error: {e}")
# Return a simple error image
error_img = Image.new('RGB', (IMG_WIDTH, IMG_HEIGHT), color=(255, 255, 255))
error_draw = ImageDraw.Draw(error_img)
try:
error_draw.text((PADDING_X, PADDING_Y), f"Error creating page {page_num}", font=font, fill=(255, 0, 0))
except:
pass
return error_img
# Process content with page limit
for item in all_lines_and_breaks:
if img_count >= MAX_PAGES:
break
is_break = item is None
item_height = PARAGRAPH_SPACING if is_break else line_height
if y_text + item_height > IMG_HEIGHT - PADDING_Y:
img_count += 1
create_image_page(page_content, img_count)
page_content = [item]
y_text = PADDING_Y + item_height + (0 if is_break else LINE_SPACING)
else:
page_content.append(item)
y_text += item_height + (0 if is_break else LINE_SPACING)
if page_content and img_count < MAX_PAGES:
img_count += 1
create_image_page(page_content, img_count)
# Memory cleanup
gc.collect()
return generated_images[:MAX_PAGES] # Ensure we don't exceed limit
except Exception as e:
print(f"Critical error in text_to_images_mcp: {e}")
print(traceback.format_exc())
# Return a simple error image
try:
error_img = Image.new('RGB', (1080, 1080), color=(255, 255, 255))
error_draw = ImageDraw.Draw(error_img)
error_draw.text((80, 80), f"Error processing text: {str(e)[:100]}", fill=(255, 0, 0))
return [error_img]
except:
return []
# Sample text for demonstration
sample_text = """In the heart of a bustling city, there lived a clockmaker named Alistair. His shop, a quaint corner of tranquility amidst the urban chaos, was filled with the gentle ticking of countless timepieces. Each clock was a masterpiece, a testament to his dedication and skill. But Alistair held a secret. One of his clocks, an old grandfather clock in the corner, did not just tell time. It told stories.
Every midnight, as the city slept, the clock would chime, not with bells, but with whispers of forgotten tales. Stories of ancient kings, lost love, and adventures in lands woven from starlight and dreams. Alistair would sit by the fire, listening, his heart filled with the magic of the past. He was the keeper of time, and in turn, time had made him its confidant.
One day, a young girl with eyes as curious as a cat's wandered into his shop. She wasn't interested in the shiny new watches but was drawn to the grandfather clock. "What's its story?" she asked, her voice soft. Alistair smiled, for he knew he had found the next guardian of the stories. The legacy of the whispering clock would live on."""
def create_app():
"""Create and configure the Gradio app with better error handling."""
# Create the demo with enhanced error handling
demo = gr.Interface(
fn=text_to_images_mcp,
inputs=[
gr.Textbox(
value=sample_text,
lines=10,
max_lines=20,
label="Text Content",
placeholder="Enter your long-form text here (max 10,000 characters)..."
),
gr.Dropdown(
choices=['plain', 'lines', 'dots', 'grid'],
value='lines',
label="Background Style",
info="Choose the background style for your images"
),
gr.Textbox(
value="",
label="Custom Font Path (Optional)",
placeholder="/path/to/your/font.ttf",
info="Leave empty to use system default font"
)
],
outputs=[
gr.Gallery(
label="Generated Images (Max 10 pages)",
show_label=True,
elem_id="gallery",
columns=2,
rows=2,
object_fit="contain",
height="auto",
show_download_button=True
)
],
title="πŸ“– Text to Images Converter (MCP Compatible)",
description="Transform long-form text into a series of attractive, readable images. Simply paste your text, choose a background style, and preview the generated images. You can download individual images by clicking on them. Limited to 10,000 characters and 10 pages for optimal performance. This app functions as an MCP server for LLM integration.",
examples=[
[sample_text, 'lines', ''],
[sample_text, 'dots', ''],
[sample_text, 'grid', ''],
[sample_text, 'plain', '']
],
cache_examples=False,
show_api=True
)
return demo
if __name__ == "__main__":
# Set up environment for MCP server
os.environ["GRADIO_MCP_SERVER"] = "True"
# Create the app
demo = create_app()
# Launch with multiple fallback strategies
launch_configs = [
# Strategy 1: Full MCP server with explicit settings
{
"mcp_server": True,
"server_name": "0.0.0.0",
"server_port": 7860,
"share": False,
"debug": False,
"show_error": False,
"quiet": True
},
# Strategy 2: Environment variable only
{
"server_name": "0.0.0.0",
"server_port": 7860,
"share": False,
"debug": False,
"show_error": False,
"quiet": True
},
# Strategy 3: Minimal configuration
{
"server_name": "0.0.0.0",
"server_port": 7860,
"share": False
},
# Strategy 4: Default configuration
{}
]
launched = False
for i, config in enumerate(launch_configs):
try:
print(f"Attempting launch strategy {i+1}...")
demo.launch(**config)
launched = True
print(f"Successfully launched with strategy {i+1}")
break
except Exception as e:
print(f"Launch strategy {i+1} failed: {e}")
if i < len(launch_configs) - 1:
print("Trying next strategy...")
continue
if not launched:
print("All launch strategies failed. Please check your environment and dependencies.")
print("You may need to install gradio[mcp] or update your gradio version.")