T2I / app.py
VirtualOasis's picture
Update app.py
3c7dd24 verified
raw
history blame
9.01 kB
import os
import gradio as gr
from PIL import Image, ImageDraw, ImageFont
import base64
import io
# --- Image Generation Logic ---
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."""
for x in range(0, width, spacing):
draw.line([(x, 0), (x, height)], fill=color, width=2)
for y in range(0, height, spacing):
draw.line([(0, y), (width, y)], fill=color, width=2)
def text_to_images_and_base64_generator(text_content, style, font_size, bg_color_name, aspect_ratio_str, font_choice):
"""
Converts text into images, returning both PIL objects for UI preview
and base64 strings for an API/MCP client, with customizable options.
Args:
text_content (str): The text to be converted.
style (str): The background style ('plain', 'lines', 'dots', 'grid').
font_size (int): The font size for the text.
bg_color_name (str): The name of the background color.
aspect_ratio_str (str): The string representing the desired aspect ratio.
font_choice (str): The name of the font family to use.
Returns:
tuple: A tuple containing (list of PIL.Image.Image, list of base64 strings).
"""
if not text_content or not text_content.strip():
gr.Warning("Input text is empty. Please enter some text to generate images.")
return [], []
# --- Aspect Ratio, Color, and Font Mapping ---
aspect_ratio_map = {
"1:1 (Square)": (1080, 1080),
"16:9 (Widescreen)": (1080, 608),
"9:16 (Vertical)": (1080, 1920),
"3:4 (Standard)": (1080, 1440),
"4:3 (Standard)": (1080, 810)
}
color_map = {
"White": (255, 255, 255),
"Light Gray": (240, 240, 240),
"Beige": (245, 245, 220),
"Pale Blue": (220, 230, 245)
}
font_map = {
"Arial": ["arial.ttf", "Arial.ttf"],
"Times New Roman": ["times.ttf", "Times New Roman.ttf"],
"Courier New": ["cour.ttf", "Courier New.ttf"],
"Verdana": ["verdana.ttf", "Verdana.ttf"],
"Georgia": ["georgia.ttf", "Georgia.ttf"],
"DejaVu Sans": ["DejaVuSans.ttf"] # Good fallback
}
IMG_WIDTH, IMG_HEIGHT = aspect_ratio_map.get(aspect_ratio_str, (1080, 1080))
BACKGROUND_COLOR = color_map.get(bg_color_name, (255, 255, 255))
# --- Configuration ---
TEXT_COLOR = (10, 10, 10)
STYLE_COLOR = (225, 225, 225)
PADDING_X = 80
PADDING_Y = 80
LINE_SPACING = 20
# --- Font Loading ---
font = None
selected_font_files = font_map.get(font_choice, [])
font_paths_to_try = selected_font_files + [
"Arial.ttf", "arial.ttf", "DejaVuSans.ttf",
"/System/Library/Fonts/Supplemental/Arial.ttf",
"/usr/share/fonts/truetype/liberation/LiberationSans-Regular.ttf",
"/usr/share/fonts/truetype/dejavu/DejaVuSans.ttf"
]
for f_path in font_paths_to_try:
try:
font = ImageFont.truetype(f_path, font_size)
break
except IOError:
continue
if not font:
font = ImageFont.load_default()
gr.Warning(f"Could not find '{font_choice}' or fallback fonts. Using default.")
# --- Text Wrapping ---
drawable_width = IMG_WIDTH - 2 * PADDING_X
paragraphs = [p.strip() for p in text_content.strip().split('\n') if p.strip()]
all_lines_and_breaks = []
for i, paragraph in enumerate(paragraphs):
words = paragraph.split()
current_line = ""
for word in words:
if 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
if font.getlength(current_line + " " + word) <= drawable_width:
current_line += " " + word
else:
all_lines_and_breaks.append(current_line.strip())
current_line = word
all_lines_and_breaks.append(current_line.strip())
if i < len(paragraphs) - 1:
all_lines_and_breaks.append(None)
# --- Image Generation & Encoding ---
generated_images = []
generated_base64_list = []
page_content = []
y_text = PADDING_Y
try:
line_height = font.getbbox("A")[3] - font.getbbox("A")[1]
except AttributeError:
bbox = font.getmask("A").getbbox()
line_height = bbox[3] - bbox[1] if bbox else 12
PARAGRAPH_SPACING = line_height
def create_and_encode_page(content):
"""Helper to create a PIL image, draw text, and encode to base64."""
img = Image.new('RGB', (IMG_WIDTH, IMG_HEIGHT), color=BACKGROUND_COLOR)
draw = ImageDraw.Draw(img)
if style == 'lines':
line_style_spacing = line_height + LINE_SPACING
draw_horizontal_lines(draw, IMG_WIDTH, IMG_HEIGHT, spacing=int(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)
current_y = PADDING_Y
for page_item in content:
if page_item is not None:
draw.text((PADDING_X, current_y), page_item, font=font, fill=TEXT_COLOR)
current_y += line_height + LINE_SPACING
else:
current_y += PARAGRAPH_SPACING
generated_images.append(img)
buffered = io.BytesIO()
img.save(buffered, format="PNG")
img_base64 = base64.b64encode(buffered.getvalue()).decode("utf-8")
generated_base64_list.append(img_base64)
# --- Page Creation Loop ---
for item in all_lines_and_breaks:
is_break = item is None
item_height = PARAGRAPH_SPACING if is_break else line_height
if y_text + item_height > IMG_HEIGHT - PADDING_Y:
create_and_encode_page(page_content)
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:
create_and_encode_page(page_content)
return generated_images, generated_base64_list
# --- Gradio Interface ---
example_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.
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 in the corner. "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."""
demo = gr.Interface(
fn=text_to_images_and_base64_generator,
inputs=[
gr.Textbox(lines=10, label="Text Content", placeholder="Paste your long-form text here...", value=example_text),
gr.Radio(['lines', 'dots', 'grid', 'plain'], label="Background Style", value='lines'),
gr.Slider(24, 72, value=48, step=2, label="Font Size"),
gr.Radio(["White", "Light Gray", "Beige", "Pale Blue"], label="Background Color", value="White"),
gr.Radio(["1:1 (Square)", "16:9 (Widescreen)", "9:16 (Vertical)", "4:3 (Standard)"], label="Aspect Ratio", value="1:1 (Square)"),
gr.Dropdown(["Arial", "Times New Roman", "Courier New", "Verdana", "Georgia", "DejaVu Sans"], label="Font Family", value="Arial")
],
outputs=[
gr.Gallery(label="Generated Images (Preview)", show_label=True, preview=True),
gr.JSON(label="API Output (Base64 Strings)")
],
title="Text-to-Image Server",
description="Transforms long-form text into images. A gallery is shown for UI preview, and a JSON list of base64 strings is returned for API/MCP clients.",
allow_flagging="never"
)
# --- Main Execution ---
if __name__ == "__main__":
demo.launch(mcp_server=True)