bilder-builder / app.py
Sebastiankay's picture
8. Sept. 2024, 03:55
8be0fbe
raw
history blame
12.8 kB
import os, sys, json, re, time, base64, random, requests, shutil
import gradio as gr
from datetime import datetime
import urllib.parse
from groq import Groq
from exif import Image
from PIL import Image as PILImage, ExifTags as PILExifTags
import colorsys
# MARK: INIT
MAX_SEED = 9999
MAX_IMAGE_SIZE = 2048
GROQ_APIKEY_PROMPTENHANCE = os.environ["GROQ_APIKEY_PROMPTENHANCE"]
# delete cache folder if exist and create new
CACHE_DIR = os.path.join(os.path.dirname(__file__), "cache")
IMAGE_DIR = os.path.join(CACHE_DIR, "images")
if os.path.exists(CACHE_DIR):
shutil.rmtree(CACHE_DIR)
os.makedirs(CACHE_DIR)
os.makedirs(IMAGE_DIR)
RES = os.path.join(os.path.dirname(__file__), "_res")
gr.set_static_paths(paths=["_res/assets/"])
fontawesome_css = RES + "/fontawesome.all.min.css"
catppuccin_css = RES + "/bootstrap-catppuccin-mocha.min.css"
custom_css = RES + "/_custom.css"
custom_js = RES + "/_custom.js"
custom_head = f"""
<link rel="stylesheet" href="{fontawesome_css}">
<link rel="stylesheet" href="{catppuccin_css}">
"""
theme = gr.themes.Soft(
primary_hue="cyan",
radius_size="sm",
neutral_hue=gr.themes.Color(c100="#a6adc8", c200="#9399b2", c300="#7f849c", c400="#6c7086", c50="#cdd6f4", c500="#585b70", c600="#45475a", c700="#313244", c800="#1e1e2e", c900="#181825", c950="#11111b"),
)
title = "Bilder Builder"
# MARK: READ EXIF
def read_exif(image_path):
with open(image_path, "rb") as src:
img = Image(src)
img_comment = json.loads(img.user_comment)
# checking if the key exists before removing
if "concept" in img_comment:
img_comment.pop("concept")
return img_comment
def read_image_exfi_data(image_path):
print("Imagepath:", image_path)
img_exif_make, img_exif_comment = read_exif(image_path)
return None, image_path, img_exif_comment
# MARK: GROQ PROMPT ENHANCE
def groq_enhance_process(Prompt=""):
client = Groq(api_key=GROQ_APIKEY_PROMPTENHANCE)
Prompt = "random prompt" if Prompt == "" else Prompt
SYSTEMPROMPT = os.path.join(RES, "groq_systemmessage_prompt_enhance.json")
with open(SYSTEMPROMPT, "r") as f:
SYSTEMPROMPT = json.load(f)
completion = client.chat.completions.create(
model="llama-3.1-70b-versatile",
messages=[SYSTEMPROMPT, {"role": "user", "content": Prompt}],
temperature=0.8,
max_tokens=1024,
top_p=0.9,
stream=False,
stop=None,
)
if completion.choices[0].message.content != "":
enhanced_prompt = completion.choices[0].message.content
enhanced_prompt = re.sub(r"[\.\"]+", "", enhanced_prompt)
return enhanced_prompt
def image_get_size(image_path):
img = PILImage.open(image_path)
print("Image size:", img.size)
width, height = img.size
return width, height
# MARK: DOMINANT COLOR
def image_get_dominant_color(image_path):
img = PILImage.open(image_path)
img = img.convert("RGB")
img = img.resize((100, 100), resample=0)
pixels = list(img.getdata())
# Erzeuge eine Liste mit den Häufigkeiten der Farben
colors = []
for pixel in pixels:
r, g, b = pixel
h, s, v = colorsys.rgb_to_hsv(r / 255, g / 255, b / 255)
if v > 0.5: # Filteriere hellere Farben aus
continue
if v > 0.99: # Filteriere Weiß aus
continue
colors.append((h, s, v))
# Ermittle die dominante Farbe
dominant_color = max(colors, key=lambda x: x[2])
dominant_color_rgb = colorsys.hsv_to_rgb(dominant_color[0], dominant_color[1], dominant_color[2])
dominant_color_rgb = [int(c * 255) for c in dominant_color_rgb]
dominant_color_rgb = f"rgb({dominant_color_rgb[0]}, {dominant_color_rgb[1]}, {dominant_color_rgb[2]})"
print(dominant_color_rgb)
return dominant_color_rgb
# MARK: PROCESS IMAGE
def process(Prompt, image_width, image_height, image_seed, randomize_seed):
if Prompt == "":
gr.Info("Kein Prompt angegeben, es wird ein zufälliger Prompt generiert.", duration=15)
Prompt = groq_enhance_process("random prompt")
used_seed = random.randint(0, MAX_SEED) if image_seed == 0 or randomize_seed else image_seed
used_model = "turbo" if int(image_width) > 1024 or int(image_height) > 1024 else "flux" # turbo, flux
timestamp = datetime.now().strftime("%Y-%m-%d-%H-%M-%S")
filename_prompt = re.sub(r"[^\w\s-]", "", Prompt).strip().replace(" ", "_")
filename = timestamp + "_" + filename_prompt[:100] + ".png"
file_path = os.path.join(IMAGE_DIR, filename)
encode_prompt = urllib.parse.quote(Prompt)
request_url = f"https://image.pollinations.ai/prompt/{encode_prompt}?model={used_model}&width={image_width}&height={image_height}&nologo=true&enhance=false&nofeed=true&seed={used_seed}"
print(request_url)
response = requests.get(request_url)
if response.status_code == 200:
with open(file_path, "wb") as f:
f.write(response.content)
img_exif_comment = read_exif(file_path)
img_dominant_color = image_get_dominant_color(file_path)
img_width, img_height = image_get_size(file_path)
return ({"value": file_path, "__type__": "update"}, {"value": None, "visible": False, "__type__": "update"}, img_exif_comment, {"visible": True, "__type__": "update"}, {"value": Prompt, "visible": True, "__type__": "update"}, img_width, img_height, used_seed, {"value": file_path, "visible": True, "__type__": "update"}, img_dominant_color, used_seed)
# MARK: Gradio BLOCKS UI
with gr.Blocks(theme=theme, css=custom_css, js=custom_js, title=title) as demo:
with gr.Tab("Image Generator"):
with gr.Row():
with gr.Column(min_width=420, scale=1):
with gr.Row():
placeholder_text = "[???] Generiert dir einen zufälligen Prompt.\n[STERN] optimiert deinen eignen Prompt.\n[RUN] generiert dein Bild."
text_prompt = gr.Textbox(label="Prompt", show_label=False, lines=12, max_lines=18, placeholder=placeholder_text, elem_classes="prompt-input", autofocus=True)
with gr.Row():
random_prompt_button = gr.Button("", variant="secondary", elem_id="random_prompt_btn", elem_classes="random-prompt-btn", icon="_res/assets/star_light_48.png")
enhance_prompt_button = gr.Button("", variant="secondary", elem_id="enhance_prompt_btn", elem_classes="enhance-prompt-btn", icon="_res/assets/star_light_48.png")
run_button = gr.Button("Erstellen", variant="primary", elem_id="run_btn", elem_classes="run-btn")
with gr.Row(elem_classes="image_size_selctor_wrapper"):
with gr.Column(scale=1):
with gr.Row():
image_width = gr.Number(label="Breite", minimum=256, maximum=MAX_IMAGE_SIZE, value=1024, step=128, elem_id="image_width_selector", elem_classes="image-width-selector", scale=1)
image_height = gr.Number(label="Höhe", minimum=256, maximum=MAX_IMAGE_SIZE, value=683, step=128, elem_id="image_height_selector", elem_classes="image-height-selector", scale=1)
with gr.Row():
image_ratio_buttons = gr.Radio(["16:9", "4:3", "3:2", "1:1"], value="3:2", label="Querformat", show_label=True, info="Die gängigsten Seitenverhältnissen.", interactive=True, elem_id="image_ratio_buttons", elem_classes="image-ratio-buttons", container=True, scale=2)
switch_width_height = gr.Button("", size="sm", elem_id="switch_width_height", elem_classes="switch-ratio-btn", variant="primary", scale=1)
with gr.Column():
image_seed = gr.Slider(label="Seed", info="Jeder Seed generiert ein anderes Bild mit dem selben Prompt", minimum=0, step=1, value=42, maximum=MAX_SEED)
randomize_seed = gr.Checkbox(label="Randomize seed", value=False)
with gr.Column(min_width=600, scale=3):
with gr.Row():
with gr.Column(scale=1):
with gr.Row():
output_image = gr.Image(show_label=False, height=720, min_width=320, scale=3, elem_id="output_image", type="filepath")
with gr.Column(scale=1, visible=False, elem_classes="image-info-wrapper") as image_info_wrapper:
# with gr.Group(visible=False) as image_info_wrapper:
with gr.Group():
image_informations = gr.Markdown("""## Bildinformationen""", visible=True)
# textbox_your_prompt = gr.Textbox(label="Dein Prmopt", lines=2, max_lines=4, interactive=False, show_copy_button=True)
with gr.Row(elem_classes="img-size-wrapper"):
image_info_tb_width = gr.Textbox(label="Breite", lines=1, max_lines=1, interactive=False, show_copy_button=True, elem_classes="image-info-tb-width")
image_info_tb_height = gr.Textbox(label="Höhe", lines=1, max_lines=1, interactive=False, show_copy_button=True, elem_classes="image-info-tb-height")
with gr.Row(elem_classes="img-seed-wrapper"):
image_info_tb_seed = gr.Textbox(label="Seed", lines=1, max_lines=1, interactive=False, show_copy_button=True, elem_classes="image-info-tb-seed")
image_info_tb_prompt = gr.Textbox("Bild Prompt", lines=4, max_lines=8, interactive=False, show_copy_button=True, visible=True)
image_download_button = gr.DownloadButton("Bild herunterladen", value=None, elem_classes="download-button", variant="primary", visible=False)
output_url = gr.Textbox(label="Output URL", show_label=True, interactive=False, visible=False)
outpu_image_comment = gr.Json(visible=False)
output_dominant_image_color = gr.Textbox(show_label=False, elem_id="dominant_image_color", visible=True, elem_classes="output-dominant-image-color")
def switch_image_size_values(image_width, image_height):
return image_height, image_width
def switch_image_ratio_buttons(ratio_value):
ratio_value = ratio_value.split(":")
ratio_value_new = f"{int(ratio_value[1])}:{int(ratio_value[0])}"
if ratio_value[1] > ratio_value[0]:
# Querformat
new_choises = ["16:9", "4:3", "3:2", "1:1"]
new_label = "Querformat"
elif ratio_value[1] < ratio_value[0]:
# Hochformat
new_choises = ["9:16", "3:4", "2:3", "1:1"]
new_label = "Hochformat"
return {"choices": new_choises, "value": ratio_value_new, "label": new_label, "__type__": "update"}
def calculate_ratio_values(image_ratio_buttons, image_width, image_height):
ratio_value = image_ratio_buttons.split(":")
if int(ratio_value[0]) > int(ratio_value[1]):
a = 1024
b = int(a * int(ratio_value[1]) / int(ratio_value[0]))
new_width = a
new_height = b
elif int(ratio_value[0]) < int(ratio_value[1]):
b = 1024
a = int(b * int(ratio_value[0]) / int(ratio_value[1]))
new_width = a
new_height = b
else:
new_width = 1024
new_height = 1024
return new_width, new_height
switch_width_height.click(fn=switch_image_size_values, inputs=[image_width, image_height], outputs=[image_width, image_height], show_progress="hidden")
switch_width_height.click(fn=switch_image_ratio_buttons, inputs=[image_ratio_buttons], outputs=[image_ratio_buttons], show_progress="hidden")
image_ratio_buttons.input(fn=calculate_ratio_values, inputs=[image_ratio_buttons, image_width, image_height], outputs=[image_width, image_height], show_progress="hidden")
run_button.click(fn=process, inputs=[text_prompt, image_width, image_height, image_seed, randomize_seed], outputs=[output_image, output_url, outpu_image_comment, image_informations, image_info_tb_prompt, image_info_tb_width, image_info_tb_height, image_info_tb_seed, image_download_button, output_dominant_image_color, image_seed])
enhance_prompt_button.click(fn=groq_enhance_process, inputs=[text_prompt], outputs=[text_prompt])
random_prompt_button.click(fn=groq_enhance_process, inputs=None, outputs=[text_prompt])
# MARK: Gradio LAUNCH
demo.launch()