Gopalag's picture
Update app.py
155ee81 verified
raw
history blame
10.9 kB
import gradio as gr
import numpy as np
import random
import spaces
import torch
from diffusers import DiffusionPipeline
from PIL import Image
dtype = torch.bfloat16
device = "cuda" if torch.cuda.is_available() else "cpu"
pipe = DiffusionPipeline.from_pretrained(
"black-forest-labs/FLUX.1-schnell",
torch_dtype=dtype
).to(device)
MAX_SEED = np.iinfo(np.int32).max
MAX_IMAGE_SIZE = 2048
STYLE_OPTIONS = {
"Vintage": "vintage style, retro aesthetic, aged appearance",
"Realistic": "photorealistic, detailed, true-to-life",
"Geometric": "geometric shapes, precise lines, mathematical patterns",
"Abstract": "abstract design, non-representational, artistic",
"Minimalist": "simple, clean lines, understated",
"Bohemian": "boho style, free-spirited, eclectic",
"Traditional": "classical design, timeless patterns",
"Contemporary": "modern style, current trends"
}
FABRIC_OPTIONS = {
"None": "",
"Cotton": "cotton textile texture, natural fiber appearance",
"Silk": "silk fabric texture, smooth and lustrous",
"Linen": "linen texture, natural weave pattern",
"Velvet": "velvet texture, plush surface",
"Canvas": "canvas texture, sturdy weave pattern",
"Wool": "wool texture, natural fiber appearance"
}
def enhance_prompt_for_pattern(prompt, style, fabric):
"""Add specific terms to ensure seamless, tileable patterns with style and fabric considerations."""
pattern_terms = [
"seamless pattern",
"tileable textile design",
"repeating pattern",
"high-quality fabric design",
"continuous pattern",
]
enhanced_prompt = f"{prompt}, {random.choice(pattern_terms)}"
if style and style != "None":
enhanced_prompt += f", {STYLE_OPTIONS[style]}"
if fabric and fabric != "None":
enhanced_prompt += f", {FABRIC_OPTIONS[fabric]}"
enhanced_prompt += ", suitable for textile printing, high-quality fabric design, seamless edges"
return enhanced_prompt
def add_logo(image):
"""Add logo to the bottom right corner of the image."""
try:
logo = Image.open('logo.png')
# Resize logo to be proportional to image size (e.g., 10% of image width)
logo_width = int(image.size[0] * 0.2)
logo_ratio = logo.size[1] / logo.size[0]
logo_height = int(logo_width * logo_ratio)
logo = logo.resize((logo_width, logo_height), Image.Resampling.LANCZOS)
# If logo has alpha channel, create a copy of the image to paste onto
if logo.mode == 'RGBA':
temp_img = image.copy()
# Calculate position for bottom right corner with small padding
position = (image.size[0] - logo_width - 20, image.size[1] - logo_height - 20)
temp_img.paste(logo, position, logo)
return temp_img
else:
# For non-transparent logos
temp_img = image.copy()
position = (image.size[0] - logo_width - 20, image.size[1] - logo_height - 20)
temp_img.paste(logo, position)
return temp_img
except Exception as e:
print(f"Error adding logo: {e}")
return image
def create_fabric_preview(image):
"""Create a fabric preview by tiling the pattern."""
# Create a 4x2 grid of the pattern
width, height = image.size
preview = Image.new('RGB', (width * 4, height * 2))
for y in range(2):
for x in range(4):
preview.paste(image, (x * width, y * height))
# Add logo to the preview
preview = add_logo(preview)
return preview
@spaces.GPU()
def infer(prompt, style, fabric, seed=42, randomize_seed=False, width=1024, height=1024,
num_inference_steps=4, progress=gr.Progress(track_tqdm=True)):
if randomize_seed:
seed = random.randint(0, MAX_SEED)
enhanced_prompt = enhance_prompt_for_pattern(prompt, style, fabric)
generator = torch.Generator().manual_seed(seed)
image = pipe(
prompt=enhanced_prompt,
width=width,
height=height,
num_inference_steps=num_inference_steps,
generator=generator,
guidance_scale=0.0
).images[0]
# Convert to PIL Image for processing
pil_image = image
if not isinstance(image, Image.Image):
pil_image = Image.fromarray(np.uint8(image))
# Add logo to single pattern
pattern_with_logo = add_logo(pil_image)
# Create fabric preview
fabric_preview = create_fabric_preview(pil_image)
return pattern_with_logo, fabric_preview, seed
examples = [
["geometric Art Deco shapes in gold and navy", "Geometric", "None"],
["abstract watercolor spots in pastel colors", "Abstract", "Silk"],
["traditional paisley design in earth tones", "Traditional", "Linen"],
["delicate floral motifs with small roses and leaves tileable textile design", "Vintage", "Cotton"],
["modern minimalist lines and circles", "Minimalist", "Canvas"],
]
# Enhanced CSS for better visual design and mobile responsiveness
css = """
#col-container {
margin: 0 auto;
max-width: 1200px !important;
padding: 20px;
}
.main-title {
text-align: center;
color: #2d3748;
margin-bottom: 1rem;
font-family: 'Poppins', sans-serif;
}
.subtitle {
text-align: center;
color: #4a5568;
margin-bottom: 2rem;
font-family: 'Inter', sans-serif;
font-size: 0.95rem;
line-height: 1.5;
}
.pattern-input {
border: 2px solid #e2e8f0;
border-radius: 10px;
padding: 12px !important;
margin-bottom: 1rem !important;
font-size: 1rem;
transition: all 0.3s ease;
}
.pattern-input:focus {
border-color: #4299e1;
box-shadow: 0 0 0 3px rgba(66, 153, 225, 0.1);
}
.generate-button {
background-color: #4299e1 !important;
color: white !important;
padding: 12px 24px !important;
border-radius: 8px !important;
font-weight: 600 !important;
transition: all 0.3s ease !important;
}
.generate-button:hover {
background-color: #3182ce !important;
transform: translateY(-1px);
}
.result-image {
border-radius: 12px;
box-shadow: 0 4px 6px -1px rgba(0, 0, 0, 0.1);
margin-top: 1rem;
}
.advanced-settings {
margin-top: 1.5rem;
border: 1px solid #e2e8f0;
border-radius: 10px;
padding: 1rem;
}
.examples-section {
margin-top: 2rem;
padding: 1rem;
background: #f7fafc;
border-radius: 10px;
border: none;
}
.preview-section {
margin-top: 1rem;
padding: 1rem;
background: #ffffff;
border-radius: 10px;
}
/* Mobile Responsiveness */
@media (max-width: 768px) {
#col-container {
padding: 12px;
}
.main-title {
font-size: 1.5rem;
}
.subtitle {
font-size: 0.9rem;
}
.pattern-input {
font-size: 0.9rem;
}
}
"""
with gr.Blocks(css=css, theme=gr.themes.Soft()) as demo:
with gr.Column(elem_id="col-container"):
gr.Markdown(
"""
# 🎨 Professional Textile Pattern Generator
""",
elem_classes=["main-title"]
)
gr.Markdown(
"""
Create professional-grade, seamless patterns for textile manufacturing.
Design unique patterns with style and fabric texture controls,
perfect for commercial textile production and fashion design.
""",
elem_classes=["subtitle"]
)
with gr.Row():
with gr.Column(scale=2):
prompt = gr.Text(
label="Pattern Description",
show_label=False,
max_lines=1,
placeholder="Describe your dream pattern (e.g., 'geometric Art Deco shapes in gold and navy')",
container=False,
elem_classes=["pattern-input"]
)
with gr.Column(scale=1):
style = gr.Dropdown(
choices=list(STYLE_OPTIONS.keys()),
label="Style",
value="None"
)
with gr.Column(scale=1):
fabric = gr.Dropdown(
choices=list(FABRIC_OPTIONS.keys()),
label="Fabric Texture",
value="None"
)
with gr.Column(scale=0.5):
run_button = gr.Button(
"✨ Generate",
elem_classes=["generate-button"]
)
with gr.Row():
with gr.Column():
pattern = gr.Image(
label="Generated Pattern",
show_label=True,
elem_classes=["result-image"]
)
with gr.Column():
preview = gr.Image(
label="Fabric Preview",
show_label=True,
elem_classes=["result-image"]
)
with gr.Accordion("🔧 Advanced Settings", open=False):
with gr.Group(elem_classes=["advanced-settings"]):
seed = gr.Slider(
label="Pattern Seed",
minimum=0,
maximum=MAX_SEED,
step=1,
value=0,
)
randomize_seed = gr.Checkbox(
label="Randomize Pattern",
value=True
)
with gr.Row():
width = gr.Slider(
label="Width",
minimum=256,
maximum=MAX_IMAGE_SIZE,
step=32,
value=1024,
)
height = gr.Slider(
label="Height",
minimum=256,
maximum=MAX_IMAGE_SIZE,
step=32,
value=1024,
)
num_inference_steps = gr.Slider(
label="Generation Quality (Steps)",
minimum=1,
maximum=50,
step=1,
value=4,
)
with gr.Group(elem_classes=["examples-section"]):
gr.Examples(
examples=examples,
fn=infer,
inputs=[prompt, style, fabric],
outputs=[pattern, preview, seed],
cache_examples=True
)
gr.on(
triggers=[run_button.click, prompt.submit],
fn=infer,
inputs=[prompt, style, fabric, seed, randomize_seed, width, height, num_inference_steps],
outputs=[pattern, preview, seed]
)
demo.launch()