Fai-Fuzer-v0.3 / app.py
salso's picture
Update app.py
5dcbb7c verified
raw
history blame contribute delete
No virus
23.6 kB
#version 2.1
import os
import math
import gradio as gr
import numpy as np
import requests
import json
import base64
from PIL import Image
from io import BytesIO
import runpod
from enum import Enum
api_key = os.getenv("FAI_API_KEY")
api = os.getenv("FAI_API_5")
api_old = os.getenv("FAI_API_old")
api_anime = os.getenv("FAI_anime") #api_anime = os.getenv("ANIME_API")
rmbgkey = os.getenv("RMBGKEY")
#fore, prompt, intensity, mode, refprompt, isrmbg
#erased this prompt
#"A car, on a high-tech city street at night, surrounded by neon lights and holographic billboards, with sleek skyscrapers and flying cars in the background, a glowing cyber-road beneath",
outputs_list = [["examples/out1.png"],["examples/out3.png"], ["examples/out4.png"],["examples/out5.png"],["examples/out6.png"],["examples/out7.png"],["examples/out2.png"]]
prompt_list = ["A Perfume Bottle, resting on a wooden table, surrounded by lavender leaves with droplets of morning dew glistening, in a serene garden, under the soft glow of a twilight sky, encircled by delicate purple petals",
"A PAir of Shoes, resting on a sleek white platform under a bright spotlight, with subtle pastel-colored reflections and delicate flower petals scattered around, creating a cheerful and elegant setting",
"A Perfume bottle, perched delicately on a rock by the ocean, with a breathtaking sunset casting a golden glow over the waves",
"A Bottle, placed on a rustic wooden table, viewed from a slight above angle, overlooking a serene beach, with seashells and driftwood artfully scattered around, bathed in the warm glow of the setting sun",
"A Sofa, In a snug, inviting room, a large sofa accompanied by plush cushions, a thick rug, book-filled shelves, potted plants, framed art, a coffee table with candles, a floor lamp, and gentle, warm lighting",
"Headphones, hovering gracefully against a swirling pastel galaxy, accented by glimmers of stardust, ethereal light beams, and floating musical notes, under a dreamy aurora borealis, with no visible wires"
]
example_list = [
[
"examples/in5.png",
"A Perfume bottle, perched delicately on a rock ⁄by the ocean, with a breathtaking sunset casting a golden glow over the waves",
3.5,
"full",
"(( A transparent perfume Bottle with a black cap )), and a yellow liquide, sunset, transparent",
False,
"examples/out5.png"
],
[
"examples/in6.png",
"A Sofa, In a snug, inviting room, a large sofa accompanied by plush cushions, a thick rug, book-filled shelves, potted plants, framed art, a coffee table with candles, a floor lamp, and gentle, warm lighting",
3.5,
"full",
"A light grey Sofa, studio light, shadows",
False,
"examples/out6.png"
],
#[
# "examples/in7.png",
# "A car, on a high-tech city street at night, surrounded by neon lights and holographic billboards, with sleek skyscrapers and flying cars in the background, a glowing cyber-road beneath",
# 3.0,
# "full",
# "(( A light grey Car )), vibrant, reflections",
# False,
# "examples/out7.png"
#],
[
"examples/in1.png",
"A Perfume Bottle, resting on a wooden table, surrounded by lavender leaves with droplets of morning dew glistening, in a serene garden, under the soft glow of a twilight sky, encircled by delicate purple petals",
3.0,
"full",
"A perfume Bottle with a purple liquide, studio light",
False,
"examples/out1.png"
],
[
"examples/in2.png",
"Headphones, hovering gracefully against a swirling pastel galaxy, accented by glimmers of stardust, ethereal light beams, and floating musical notes, under a dreamy aurora borealis, with no visible wires",
3.50,
"full",
"headphones, vibrant, colorful, ",
False,
"examples/out2.png"
],
[
"examples/in3.png",
"A PAir of Shoes, resting on a sleek white platform under a bright spotlight, with subtle pastel-colored reflections and delicate flower petals scattered around, creating a cheerful and elegant setting",
3.5,
"full",
"(( A Pair of Brown Shoes )), vibrant, Shadow",
False,
"examples/out3.png"
],
[
"examples/in4.png",
"A Bottle, placed on a rustic wooden table, viewed from a slight above angle, overlooking a serene beach, with seashells and driftwood artfully scattered around, bathed in the warm glow of the setting sun",
3.50,
"full",
"A transparent Bottle with a black cap and a semi transparent liquide, Sunset",
False,
"examples/out4.png"
],
[
"examples/anime_1in_1.png",
"A man with spiky standing on a cliff overlooking a dystopian city at dusk",
2.5,
"full",
"Anime style, intense expression, neon lights",
False,
"examples/anime_1out.png"
],
[
"examples/anime_2in_1.png",
"A young man standing in front of a park, on a sunny day",
2.5,
"full",
"((a man with a green shirt)), Anime style, stoic, serene atmosphere, bright sky, sunny ",
False,
"examples/anime_5out.png"
],
[
"examples/anime_3in.png",
"A woman , sitting in an outdoor restaurant, beautiful cityscape",
2.5,
"full",
"Anime style, joyful expression",
False,
"examples/anime_3out.webp"
],
[
"examples/anime_4in.png",
"A woman , standing in front of mountains and sea view",
2.5,
"full",
"Anime style, detailed background",
False,
"examples/anime_4out.webp"
]
]
def rmbg(pil_image):
# Convert PIL image to bytes
image_bytes = BytesIO()
pil_image.save(image_bytes, format='PNG')
image_bytes.seek(0)
# Send the image to the remove.bg API
response = requests.post(
'https://api.remove.bg/v1.0/removebg',
files={'image_file': ('filename.png', image_bytes, 'image/png')},
data={'size': 'auto'},
headers={'X-Api-Key': rmbgkey}
)
if response.status_code == 200:
# Convert the bytes response to a PIL image
result_image = Image.open(BytesIO(response.content))
return result_image
else:
return None
def image_to_base64(image):
# Open the image file
with image:
# Create a buffer to hold the binary data
buffered = BytesIO()
# Save the image in its original format to the buffer
#print(image.format)
image.save(buffered, format="PNG")
# Get the byte data from the buffer
binary_image_data = buffered.getvalue()
# Encode the binary data to a base64 string
base64_image = base64.b64encode(binary_image_data).decode("utf-8")
return base64_image
def create_square_image(image):
"""
Create a new square image with the side length equal to the largest dimension
of the original image and paste the original image at the center on a transparent canvas.
:param image: A PIL image.
:return: A new square PIL image.
"""
original_width, original_height = image.size
new_side_length = max(original_width, original_height)
# Create a new square image with a transparent background
new_image = Image.new("RGBA", (new_side_length, new_side_length), (255, 255, 255, 0))
# Calculate the position to paste the original image on the new square canvas
paste_x = (new_side_length - original_width) // 2
paste_y = (new_side_length - original_height) // 2
# Paste the original image onto the new square canvas using the alpha channel as a mask
new_image.paste(image, (paste_x, paste_y), image)
return new_image
def process(data, api, api_key):
runpod.api_key = api_key
input_payload = {"input": data }
try:
endpoint = runpod.Endpoint(api)
run_request = endpoint.run(input_payload)
# Initial check without blocking, useful for quick tasks
status = run_request.status()
print(f"Initial job status: {status}")
if status=="IN_QUEUE":
gr.Info("Queued 🚢🚢🚢🚢!", duration=35)
if status != "COMPLETED":
# Polling with timeout for long-running tasks
output = run_request.output(timeout=120)
else:
output = run_request.output()
print(f"Job output: {output}")
except Exception as e:
print(f"An error occurred: {e}")
status = run_request.status()
if status=="FAILED":
raise gr.Error(f"An error occured πŸ’₯! {e}", duration=5)
if status=="TIMED_OUT":
raise gr.Error("Sorry we could not secure a worker for you ⏳! Try again", duration=5)
image_data = output['image']
# Decode the Base64 string
image_bytes = base64.b64decode(image_data)
# Convert binary data to image
image = Image.open(BytesIO(image_bytes))
return image
def resize_to_fit(max_size, original_size):
"""
Calculate the new size for an image to fit within max_size while maintaining the aspect ratio.
:param max_size: Maximum allowed size as a tuple (width, height).
:param original_size: Original size of the image as a tuple (width, height).
:return: New size as a tuple (new_width, new_height) that fits within max_size while maintaining the aspect ratio.
"""
original_width, original_height = original_size
max_width, max_height = max_size
# Calculate the scaling factor to maintain aspect ratio
width_ratio = max_width / original_width
height_ratio = max_height / original_height
scaling_factor = min(width_ratio, height_ratio)
# Calculate the new size while maintaining the aspect ratio
new_width = int(original_width * scaling_factor)
new_height = int(original_height * scaling_factor)
return new_width, new_height
def process_generate(fore, back, prompt, intensity, mode, refprompt, isrmbg, model_type, steps, control_strength):
max_size = (1500,1500)
if prompt in prompt_list:
max_size = (1500,1500)
if isrmbg:
try:
rmbgfore = rmbg(fore)
if rmbgfore is not None:
fore = rmbgfore.convert("RGBA")
print(f"Background removed!")
except:
pass
#
if fore is None:
print(fore)
# Select the endpoint based on the model type
if model_type == "anime":
api_endpoint = api_anime #to change to api_anime
elif model_type == "realistic":
fore = create_square_image(fore)
api_endpoint = api_old
else:
raise ValueError("Invalid model type selected.")
size = fore.size
image_width = size[0]
image_height = size[1]
gr.Warning(f"ℹ️ The input image size is: {size} ")
if size[0]*size[1]<=(512*512):
gr.Warning("ℹ️ The input image resolution is low, it might lead to some deformation!")
if size[0]*size[1]>(max_size[0]*max_size[1]):
gr.Warning("ℹ️ The input image size is too big, I will lower it!")
image_width, image_height = resize_to_fit((768,768), (image_width, image_height))
fore.resize(max_size)
if back is not None:
fore = Image.alpha_composite(back.resize(fore.size).convert('RGBA'), fore.convert('RGBA'))
print(f"Background Merged!")
mode = "refiner"
forestr = image_to_base64(fore.convert("RGBA"))
data = {
"foreground_image64": forestr,
"prompt" : prompt,
"mode" : mode,
"intensity" : float(intensity),
"width" : 1000,
"height" : 1000,
"refprompt" : refprompt,
"first_stage_steps" : int(steps),
"first_stage_strength" : float(control_strength),
"second_stage_steps" : 20
}
#print(f"DATA: {data}")
'''
data = {
"foreground_image64": forestr,
"prompt" : "There is Perfume, nestled on a crystalline cliff of glistening snow, under a celestial night sky adorned with constellations and swirling galaxies, framed by ethereal, blue flames that dance gracefully in the icy air",
"mode" : "full", #refiner, full
"intensity" : 3.0,
"width" : 1000,
"height" : 1000,
"refprompt" : " transparent glass "
}
'''
image = process(data, api_endpoint, api_key)
return image
def update_value(val):
return val
class Stage(Enum):
FIRST_STAGE = "first-stage"
SECOND_STAGE = "refiner"
FULL = "full"
css="""#disp_image {
text-align: center;
}
#share-btn-container {
padding-left: 0.5rem !important;
padding-right: 0.5rem !important;
background-color: #000000;
justify-content: center;
align-items: center;
border-radius: 9999px !important;
max-width: 13rem;
margin-left: auto;
}
#share-btn-container > div {
flex-direction: row;
background: black;
align-items: center;
}
#share-btn-container:hover {
background-color: #060606;
}
#share-btn {
all: initial;
color: #ffffff;
font-weight: 600;
cursor: pointer;
font-family: 'IBM Plex Sans', sans-serif;
margin-left: 0.5rem !important;
padding-top: 0.5rem !important;
padding-bottom: 0.5rem !important;
right: 0;
}
#share-btn * {
all: unset;
}
#share-btn-container div:nth-child(-n+2) {
width: auto !important;
min-height: 0px !important;
}
#share-btn-container .wrap {
display: none !important;
}
#share-btn-container.hidden {
display: none !important;
}
#duplicate-button {
margin-left: auto;
color: #fff;
background: #1565c0;
}
body {
font-family: Arial, sans-serif;
background-color: #f4f4f9;
margin: 0;
padding: 0;
min-height: 100vh;
color: #333;
}
.custom-button {
background: linear-gradient(rgba(0, 0, 0, 1) 0%, rgba(0, 0, 0, 0.8) 100%);
color: transparent;
font-size: 30px;
color: transparent;
padding: 10px 20px;
border: none;
border-radius: 5px;
cursor: pointer;
transition: opacity 0.3s ease;
position: relative;
}
.custom-button:hover {
opacity: 0.8;
}
.custom-button::before {
content: 'Generate';
background: linear-gradient(90deg, rgb(29, 161, 255) 0%, rgb(163, 44, 255) 56%, rgb(251, 156, 255) 100%);
-webkit-background-clip: text;
-webkit-text-fill-color: transparent;
position: absolute;
top: 0;
left: 0;
right: 0;
bottom: 0;
display: flex;
align-items: center;
justify-content: center;
z-index: 1;
font-weight: bold;
}
.custom-title {
font-size: 36px;
background: linear-gradient(90deg, rgb(29, 161, 255) 0%, rgb(163, 44, 255) 56%, rgb(251, 156, 255) 100%) text;
-webkit-background-clip: text;
-webkit-text-fill-color: transparent;
font-weight: bold;
text-align: center;
margin-bottom: 20px;
}
h1 {
color: #222222;
}
a {
color: #4a90e2;
text-decoration: none;
font-weight: bold;
}
a:hover {
text-decoration: underline;
}
.emoji {
font-size: 1.5em;
}
@media (max-width: 768px) {
.custom-button {
font-size: 20px;
padding: 8px 16px;
}
.custom-title {
font-size: 28px;
}
#disp_image, .container {
padding: 0 1rem;
}
h1 {
font-size: 24px;
}
}
@media (max-width: 480px) {
.custom-button {
font-size: 16px;
padding: 6px 12px;
}
.custom-title {
font-size: 22px;
}
.container {
padding: 0 0.5rem;
}
h1 {
font-size: 20px;
}
#share-btn-container {
max-width: 100%;
padding: 0.5rem;
}
}
"""
block = gr.Blocks(css=css, title="## FAI Fuzer").queue(default_concurrency_limit=12)
with block:
gr.HTML('''
<div class="container">
<center>
<h1 class="custom-title">
FAI Fuzer medium v0.3: Empower your AI Image Generation with Full Control
<span class="emoji"></span>
</h1>
</center>
<h2>Welcome to the Updated Version of FAI Fuzer!</h2>
<p>This is FAI Fuzer medium v0.3, an updated version of our previously released Fuzer v0.1. In this version, we've introduced several enhancements, including the addition of a new anime model to broaden your creative possibilities.</p>
<p>We appreciate the feedback and support from our community and are excited to bring these new features to you.</p>
<p><strong>What's New in v0.3:</strong></p>
<ul>
<li>πŸš€ Introduction of an Anime Model: Now you can choose between "realistic" and "anime" styles for your images.</li>
<li>🎨 Enhanced Control: Improved background blending and consistency features.</li>
<li>πŸ“ˆ Performance Improvements: Faster processing and improved UI responsiveness.</li>
</ul>
<p>Please check us out on <strong>Twitter</strong>:
<a href="https://x.com/FotographerAI">Fotographer AI</a> and
<strong>Discord</strong>: <a href="https://discord.gg/b9RuYQ3F8k">https://discord.gg/b9RuYQ3F8k</a></p>
<h2>Step-by-Step Instructions</h2>
<p>Follow these instructions to control the generation of backgrounds while keeping the foreground's shape and style consistent:</p>
<ul>
<li>
<span class="emoji">πŸ“</span>
<strong>Step 1: Describe the Background</strong>
<p>Start by providing a detailed description of the background you want to create.</p>
<div class="example">
<p><strong>Example:</strong> "A Perfume Bottle nestled on a crystalline cliff of glistening snow, overlooking a serene, moonlit valley."</p>
</div>
</li>
<li>
<span class="emoji">πŸ’‘</span>
<strong>Step 2: Describe the Foreground</strong>
<p>Next, describe the texture, lighting, and style of the foreground element.</p>
<div class="example">
<p><strong>Example:</strong> "A transparent glass perfume bottle, vibrant, sunset lighting reflecting off its surface."</p>
</div>
</li>
<li>
<span class="emoji">🎚️</span>
<strong>Step 3: Adjust the Intensity</strong>
<p>Decide how much change you want to apply to the image. Adjust the intensity to balance between keeping consistency and introducing new elements.</p>
</li>
</ul>
<p><strong>Note: It usually takes around 20 seconds to generate an image. However, the time can extend up to 40-50 seconds when there is a queue.</strong></p>
</div>
''')
gr.HTML("""
<center><h2><a href="https://fotographer.ai/api/top">πŸ”— Check out our API!</a></h2></center>""")
with gr.Row():
gr.Markdown("### FAI Fuzer medium: Composite Photography in various styles in less than 2 minutes!")
with gr.Row():
fore = gr.Image(image_mode='RGBA', type="pil", label="Foreground Image", height=400, width=250, min_width=250)
back = gr.Image(image_mode='RGBA', type="pil", label="Background Image", height=400, width=250, min_width=250)
# with gr.Column():
result_gallery = gr.Image(label='Output', min_width=400) #gr.Gallery(height=400, object_fit='contain', label='Outputs')
with gr.Row():
prompt = gr.Textbox(label="Prompt", min_width=400)
# with gr.Column():
refprompt = gr.Textbox(label="Refiner Prompt", min_width=400)
with gr.Row():
with gr.Column(min_width=400):
mode = gr.Radio(choices=[e.value for e in Stage],
value=Stage.FULL.value,
label="Generation Mode", type='value', min_width=400)
mode.change(fn=update_value, inputs=mode, outputs=mode)
model_type = gr.Dropdown(choices=["realistic","anime"], label="Model Type", value="realistic", min_width=400) # Adding the Dropdown for model selection
with gr.Column(min_width=400):
# Adding Advanced Settings for the number of steps and control strength
with gr.Accordion("Advanced Settings", open=True, ):
steps = gr.Slider(label="Number of Steps", minimum=10, maximum=100, value=30, step=1)
control_strength = gr.Slider(label="Control Strength", minimum=0.1, maximum=1.0, value=0.5, step=0.1)
with gr.Column(min_width=400):
gr.HTML('''
<div class="container">
<h1>πŸš€ For more freedom of usage, check out our API</h1>
<p>πŸ‘€ You can test with free credits:</p>
<h2><a href="https://fotographer.ai/api/top">πŸ”— API Dashboard</a></h2>
</div>
''')
with gr.Row():
intensity = gr.Slider(label="Refiner Strength", minimum=1.0, maximum=7.0, value=3.0, step=0.5)
intensity.change(fn=update_value, inputs=intensity, outputs=intensity)
isrmbg = gr.Checkbox(label="Remove Background")
isrmbg.change(fn=update_value, inputs=isrmbg, outputs=isrmbg)
generate_button = gr.Button(value="Generate", elem_classes="custom-button")
gr.HTML('''
<div class="container">
<h2>Features:</h2>
<ul>
<li>πŸ–ΌοΈ <strong>Consistent Foreground Style:</strong> Keep the foreground's shape and style intact across various backgrounds.</li>
<li>πŸŒ„ <strong>Seamless Blending:</strong> Effortlessly blend foreground elements with any background for a natural look.</li>
<li>🎨 <strong>Custom Background Control:</strong> Choose or generate backgrounds that perfectly match your creative vision.</li>
<li>βš™οΈ <strong>Easy Integration:</strong> Integrates smoothly with your existing workflow and tools.</li>
</ul>
</div>
''')
gr.HTML("""
<center><h2><a href="https://fotographer.ai/">πŸ”— Check Out our other Projects Here!</a></h2></center>""")
with gr.Row():
dummy_image_for_outputs = gr.Image(visible=False, label='Result')
gr.Examples(
fn=lambda *args: [args[-1]],
examples=example_list,
inputs=[
fore, prompt, intensity, mode, refprompt, isrmbg, dummy_image_for_outputs
]
)
'''
with gr.Column():
dummy_image_for_outputs = gr.Image(visible=False, label='Result')
gr.Examples(
examples=outputs_list,
inputs=[dummy_image_for_outputs],
)
'''
ins = [fore, back, prompt, intensity, mode, refprompt, isrmbg, model_type, steps, control_strength]
generate_button.click(fn=process_generate, inputs=ins, outputs=[result_gallery])
block.launch()