app.py

from PIL import Image
from rembg import remove
import cairosvg
import io
import numpy as np
from sklearn.cluster import KMeans
from PIL import Image, ImageDraw, ImageFont
import random

import gradio as gr

def generate(logo=None, Vtubername=""):
  if(logo==None):
        gr.Warning('Please Select Your Photo📸')

  if(Vtubername==""):
        Vtubername = "unkown"
        gr.Warning('Please Select Your Name😱')

  def extract_dominant_colors(img, num_colors=3, ignore_edges=True):
    if img.mode == 'RGBA':
          image = img.convert('RGB')
    else:
          image = img

    image = image.resize((150, 150))
    data = np.array(image)

    pixels = data.reshape(-1, 3)

    if ignore_edges:
          edge_pixels = np.concatenate([data[0, :, :], data[-1, :, :], data[:, 0, :], data[:, -1, :]], axis=0)
          edge_colors, counts = np.unique(edge_pixels, axis=0, return_counts=True)
          background_color = edge_colors[counts.argmax()]
          pixels = pixels[~np.all(pixels == background_color, axis=1)]
          if len(pixels) == 0:
            return np.array([background_color,np.array([60,60,60]),np.array([255,255,255])])
          elif len(pixels) == 1:
            return np.array([pixels[0],np.array([60,60,60]),np.array([255,255,255])])
          elif len(pixels) == 2:
            return np.array([pixels[0],pixels[1],np.array([60,60,60])])


    
    model = KMeans(n_clusters=3)
    model.fit(pixels)
    
    colors = model.cluster_centers_
    
    colors = colors.round(0).astype(int)
    
    return colors

  dominant_colors = extract_dominant_colors(logo, num_colors=3)
  template_prime_colors = {
    "black color": [0, 0, 0],
    "white": [255, 255, 255],
    "red": [255, 0, 0],
    "lightgreen": [0, 255, 0],
    "blue": [0, 0, 255],
    "yellow": [255, 255, 0],
    "lightblue": [0, 255, 255],
    "pink": [255, 0, 255],
    "gray": [128, 128, 128],
    "maroon": [128, 0, 0],
    "olive": [128, 128, 0],
    "green": [0, 128, 0],
    "purple": [128, 0, 128],
    "navy": [0, 0, 128],
    "orange": [255, 165, 0],
    "bluegreen": [0, 128, 128],
    "lightpurple": [128, 128, 255],
    "skyblue color": [0, 128, 255],
    "brown": [139,69,19],    
  }
  _primary_color = dominant_colors[0]

  closest_color = "black color"
  for color in template_prime_colors:
    if np.linalg.norm(np.array(template_prime_colors[color]) - _primary_color) < np.linalg.norm(np.array(template_prime_colors[closest_color]) - _primary_color):
      closest_color = color
  primary_color = closest_color
  print(primary_color)
  secondary_color=str("rgb("+str(dominant_colors[1][0])+", "+str(dominant_colors[1][1])+", "+str(dominant_colors[1][2])+")")
  third_color=str("rgb("+str(dominant_colors[2][0])+", "+str(dominant_colors[2][1])+", "+str(dominant_colors[2][2])+")")

  import requests
  import os
  sdkey = os.environ["sdkey"]
  from huggingface_hub import InferenceClient

  client = InferenceClient(model="mistralai/Mixtral-8x7B-Instruct-v0.1")

  output = client.text_generation("Make this english to Japanese Hiragana. ex. Robert->はろー HuggingFace->はぎんぐふぇいす "+Vtubername+"->")
  hiragana = ""
  for char in output:
    if '\u3040' <= char <= '\u309f':
      hiragana += char

  
  response = requests.post(
      f"https://api.stability.ai/v2beta/stable-image/generate/sd3",
      headers={
          "authorization": f"Bearer "+sdkey,
          "accept": "image/*"
      },
      files={"none": ''},
      data={
          "model": "sd3",
          "prompt": "pop sweety cute kawaii font anime title logo drawn by adobe illustorator. Logo for kids amime. The title logo text is \""+Vtubername+"\""+", The logo text color:"+primary_color + ". Single Logo only.",
          "negative_prompt": "subtitle,face, ruby text, smoke, subscript, superscript, multiple titles, character, ugly, blurry, dirty, character face, face, watermark, low res, cropped, worst quality, jpeg artifacts, , picture frame, out of frame,animal, person face, low-res, blurry, blur, out of focus, disgusting",
          "output_format": "jpeg",
      },
  )
  image = None
  if response.status_code == 200:
    image = response.content
  else:
      gr.Warning('Your message is not allowed!')
      raise Exception(str(response.json()))
  image = Image.open(io.BytesIO(response.content))
  title_logo=remove(image)





  def get_brightness(color):
    red, green, blue = color
    return (red * 0.299 + green * 0.587 + blue * 0.114) / 255

  brighter_color = secondary_color if get_brightness(dominant_colors[1]) > get_brightness(dominant_colors[2]) else third_color
  darker_cplor = secondary_color if get_brightness(dominant_colors[1]) < get_brightness(dominant_colors[2]) else third_color

  font_color=brighter_color

  font_size=100
  stroke_width=int(100*0.1)
  stroke_color=darker_cplor
  # Load the font
  font = ImageFont.truetype("oshigo.otf", size=font_size)

  japanese_text = hiragana

    # Image setup
  tile_width, tile_height = int(font_size*1.4), int(font_size*1.4)  # Size of individual tiles
  num_tiles = len(japanese_text)
  total_width = tile_width * num_tiles
  total_height = tile_height

  # Create a new blank image
  result_image = Image.new('RGBA', (total_width, total_height), (0, 0, 0, 0))
  draw = ImageDraw.Draw(result_image)

  for i, char in enumerate(japanese_text):
      # Create an image for each character with transparency
      tile_image = Image.new('RGBA', (tile_width, tile_height), (0, 0, 0, 0))
      tile_draw = ImageDraw.Draw(tile_image)
      # Calculate text position: random within the tile
      text_width, text_height = draw.textsize(char, font=font)
      x = random.randint(0, (tile_width - text_width)//1.25)
      y = random.randint(0, (tile_height - text_height)//1.25)
      # Draw text on the tile
      tile_draw.text((x, y), char, font=font, fill="white", stroke_width=stroke_width, stroke_fill=stroke_color)

      # Paste the tile into the result image
      result_image.paste(tile_image, (i * tile_width, 0), tile_image)

  # Save or display the image
  caption = result_image

  def resize_caption_to_logo(logo, caption):
      logo_width = logo.width
      target_caption_width = int(logo_width)
      logo_height = logo.height
      target_caption_height = int(logo_height)

      if caption.width > target_caption_width:
          aspect_ratio = caption.width / target_caption_width
          new_width = int(caption.width * aspect_ratio * 0.4)
          new_height = int(caption.height * aspect_ratio * 0.4)
          print("resizing")
          resized_caption = caption.resize((new_width, new_height), Image.ANTIALIAS)
          return resized_caption
      else:
          return caption
  caption = resize_caption_to_logo(logo, caption)


  center=((title_logo.width - caption.width) // 2,title_logo.height//2)
  bottom=(title_logo.width-caption.width)//2,int(title_logo.height-caption.height-100)
  lower_right=(title_logo.width-caption.width-40,int(title_logo.height-caption.height-80))
  upper_right=(title_logo.width-caption.width-40,int(caption.height+80))


  # Define the possible positions
  positions = [
      ("center", center),
      ("bottom", bottom),
      ("lower_right", lower_right),
      ("upper_right", upper_right),
  ]

  # Randomly select a position
  position, coordinates = random.choice(positions)

  # Paste the caption at the selected position
  title_logo.paste(caption, coordinates, caption)
  return title_logo

css="""
.gradio-container{
  background-color: #fff;
  background-image:
    radial-gradient(#b4f3ea 0%, transparent 30%),
    radial-gradient(#ffffcc 0%, transparent 30%);  background-size: 40px 40px;
  background-position: 0 0, 20px 20px;
  }
h1{
    font-size: 400%!important;
    background: linear-gradient(to bottom, pink, white);
    -webkit-background-clip: text;
    -webkit-text-fill-color: transparent;
    -webkit-text-stroke: 2px pink;
    -webkit-text-stroke-width: 2px;
    -webkit-text-stroke-color: pink;
}
"""



iface = gr.Interface(
    theme=gr.themes.Default(primary_hue="pink",font=[gr.themes.GoogleFont("Mochiy Pop One")]),
    css=css,
    fn=generate,
    inputs=[gr.Image(label="VTuber Photo", type="pil"), gr.Textbox(label="VTuber Name（*alphabet only!*）")],
    outputs=gr.Image(label="Generated Logo"),
    title="VTuber Logo Generator",
    description="①Upload photo you wanna VTuberize❤️ <br>② Input the name（*alphabet only!*）⭐️ <br>③Press Submit🧙",
    examples=[["image.jpeg", "gojiteji"]],
    
    allow_flagging=False
)

# Launch the interface
iface.launch(debug=True)