from diffusers import AutoencoderKL, UNet2DConditionModel, StableDiffusionPipeline, StableDiffusionImg2ImgPipeline, DPMSolverMultistepScheduler
import gradio as gr
import torch
from PIL import Image
import utils
import datetime
import time
import psutil
import random
start_time = time.time()
is_colab = utils.is_google_colab()
state = None
current_steps = 25
class Model:
def __init__(self, name, path="", prefix=""):
self.name = name
self.path = path
self.prefix = prefix
self.pipe_t2i = None
self.pipe_i2i = None
models = [
Model("Arcane", "nitrosocke/Arcane-Diffusion", "arcane style "),
Model("Dreamlike Diffusion 1.0", "dreamlike-art/dreamlike-diffusion-1.0", "dreamlikeart "),
Model("Archer", "nitrosocke/archer-diffusion", "archer style "),
Model("Anything V4", "andite/anything-v4.0", ""),
Model("Modern Disney", "nitrosocke/mo-di-diffusion", "modern disney style "),
Model("Classic Disney", "nitrosocke/classic-anim-diffusion", "classic disney style "),
Model("Loving Vincent (Van Gogh)", "dallinmackay/Van-Gogh-diffusion", "lvngvncnt "),
Model("Wavyfusion", "wavymulder/wavyfusion", "wa-vy style "),
Model("Analog Diffusion", "wavymulder/Analog-Diffusion", "analog style "),
Model("Redshift renderer (Cinema4D)", "nitrosocke/redshift-diffusion", "redshift style "),
Model("Midjourney v4 style", "prompthero/midjourney-v4-diffusion", "mdjrny-v4 style "),
Model("Waifu", "hakurei/waifu-diffusion"),
Model("Cyberpunk Anime", "DGSpitzer/Cyberpunk-Anime-Diffusion", "dgs illustration style "),
Model("Elden Ring", "nitrosocke/elden-ring-diffusion", "elden ring style "),
Model("TrinArt v2", "naclbit/trinart_stable_diffusion_v2"),
Model("Spider-Verse", "nitrosocke/spider-verse-diffusion", "spiderverse style "),
Model("Balloon Art", "Fictiverse/Stable_Diffusion_BalloonArt_Model", "BalloonArt "),
Model("Tron Legacy", "dallinmackay/Tron-Legacy-diffusion", "trnlgcy "),
Model("Pokémon", "lambdalabs/sd-pokemon-diffusers"),
Model("Pony Diffusion", "AstraliteHeart/pony-diffusion"),
Model("Robo Diffusion", "nousr/robo-diffusion"),
Model("Epic Diffusion", "johnslegers/epic-diffusion")
]
custom_model = None
if is_colab:
models.insert(0, Model("Custom model"))
custom_model = models[0]
last_mode = "txt2img"
current_model = models[1] if is_colab else models[0]
current_model_path = current_model.path
if is_colab:
pipe = StableDiffusionPipeline.from_pretrained(
current_model.path,
torch_dtype=torch.float16,
scheduler=DPMSolverMultistepScheduler.from_pretrained(current_model.path, subfolder="scheduler"),
safety_checker=lambda images, clip_input: (images, False)
)
else:
pipe = StableDiffusionPipeline.from_pretrained(
current_model.path,
torch_dtype=torch.float16,
scheduler=DPMSolverMultistepScheduler.from_pretrained(current_model.path, subfolder="scheduler")
)
if torch.cuda.is_available():
pipe = pipe.to("cuda")
pipe.enable_xformers_memory_efficient_attention()
device = "GPU 🔥" if torch.cuda.is_available() else "CPU 🥶"
def error_str(error, title="Error"):
return f"""#### {title}
{error}""" if error else ""
def update_state(new_state):
global state
state = new_state
def update_state_info(old_state):
if state and state != old_state:
return gr.update(value=state)
def custom_model_changed(path):
models[0].path = path
global current_model
current_model = models[0]
def on_model_change(model_name):
prefix = "Enter prompt. \"" + next((m.prefix for m in models if m.name == model_name), None) + "\" is prefixed automatically" if model_name != models[0].name else "Don't forget to use the custom model prefix in the prompt!"
return gr.update(visible = model_name == models[0].name), gr.update(placeholder=prefix)
def on_steps_change(steps):
global current_steps
current_steps = steps
def pipe_callback(step: int, timestep: int, latents: torch.FloatTensor):
update_state(f"{step}/{current_steps} steps")#\nTime left, sec: {timestep/100:.0f}")
def inference(model_name, prompt, guidance, steps, n_images=1, width=512, height=512, seed=0, img=None, strength=0.5, neg_prompt=""):
update_state(" ")
print(psutil.virtual_memory()) # print memory usage
global current_model
for model in models:
if model.name == model_name:
current_model = model
model_path = current_model.path
# generator = torch.Generator('cuda').manual_seed(seed) if seed != 0 else None
if seed == 0:
seed = random.randint(0, 2147483647)
generator = torch.Generator('cuda').manual_seed(seed)
try:
if img is not None:
return img_to_img(model_path, prompt, n_images, neg_prompt, img, strength, guidance, steps, width, height, generator), f"Done. Seed: {seed}"
else:
return txt_to_img(model_path, prompt, n_images, neg_prompt, guidance, steps, width, height, generator), f"Done. Seed: {seed}"
except Exception as e:
return None, error_str(e)
def load_model(model_path, mode):
global last_mode
global pipe
global current_model_path
if model_path != current_model_path or last_mode != mode:
current_model_path = model_path
update_state(f"Loading {current_model.name} {mode} model...")
model_class = StableDiffusionPipeline if mode == "txt2img" else StableDiffusionImg2ImgPipeline
if is_colab or current_model == custom_model:
pipe = model_class.from_pretrained(
current_model_path,
torch_dtype=torch.float16,
scheduler=DPMSolverMultistepScheduler.from_pretrained(current_model.path, subfolder="scheduler"),
safety_checker=lambda images, clip_input: (images, False)
)
else:
pipe = model_class.from_pretrained(
current_model_path,
torch_dtype=torch.float16,
scheduler=DPMSolverMultistepScheduler.from_pretrained(current_model.path, subfolder="scheduler")
)
if torch.cuda.is_available():
pipe = pipe.to("cuda")
pipe.enable_xformers_memory_efficient_attention()
last_mode = mode
def process_image(img, width, height):
ratio = min(height / img.height, width / img.width)
return img.resize((int(img.width * ratio), int(img.height * ratio)), Image.LANCZOS)
def inference_image(model_path, prompt, mode, *args, **kwargs):
print(f"{datetime.datetime.now()} {mode}, model: {model_path}")
global pipe
load_model(model_path, mode)
prompt = current_model.prefix + prompt
result = pipe(
prompt,
*args,
**kwargs,
callback=pipe_callback
)
return replace_nsfw_images(result)
def txt_to_img(model_path, prompt, n_images, neg_prompt, guidance, steps, width, height, generator):
return inference_image(model_path, prompt, "txt2img",
negative_prompt=neg_prompt,
num_images_per_prompt=n_images,
num_inference_steps=int(steps),
guidance_scale=guidance,
width=width,
height=height,
generator=generator
)
def img_to_img(model_path, prompt, n_images, neg_prompt, img, strength, guidance, steps, width, height, generator):
img = process_image(img, width, height)
return inference_image(model_path, prompt, "img2img",
negative_prompt=neg_prompt,
num_images_per_prompt=n_images,
image=img,
num_inference_steps=int(steps),
strength=strength,
guidance_scale=guidance,
generator=generator
)
def replace_nsfw_images(results):
if is_colab:
return results.images
for i in range(len(results.images)):
if results.nsfw_content_detected[i]:
results.images[i] = Image.open("nsfw.png")
return results.images
# css = """.finetuned-diffusion-div div{display:inline-flex;align-items:center;gap:.8rem;font-size:1.75rem}.finetuned-diffusion-div div h1{font-weight:900;margin-bottom:7px}.finetuned-diffusion-div p{margin-bottom:10px;font-size:94%}a{text-decoration:underline}.tabs{margin-top:0;margin-bottom:0}#gallery{min-height:20rem}
# """
with gr.Blocks(css="style.css") as demo:
gr.HTML(
f"""
Finetuned Diffusion
Demo for multiple fine-tuned Stable Diffusion models, trained on different styles:
Arcane, Archer, Elden Ring, Spider-Verse, Modern Disney, Classic Disney, Loving Vincent (Van Gogh), Redshift renderer (Cinema4D), Midjourney v4 style, Waifu, Pokémon, Pony Diffusion, Robo Diffusion, Cyberpunk Anime, Tron Legacy, Balloon Art + in colab notebook you can load any other Diffusers 🧨 SD model hosted on HuggingFace 🤗.
You can skip the queue and load custom models in the colab: 
Running on
{device}{(" in a
Google Colab." if is_colab else "")}
You can also duplicate this space and upgrade to gpu by going to settings:
"""
)
with gr.Row():
with gr.Column(scale=55):
with gr.Group():
model_name = gr.Dropdown(label="Model", choices=[m.name for m in models], value=current_model.name)
with gr.Box(visible=False) as custom_model_group:
custom_model_path = gr.Textbox(label="Custom model path", placeholder="Path to model, e.g. nitrosocke/Arcane-Diffusion", interactive=True)
gr.HTML("Custom models have to be downloaded first, so give it some time.
")
with gr.Row():
prompt = gr.Textbox(label="Prompt", show_label=False, max_lines=2,placeholder="Enter prompt. Style applied automatically").style(container=False)
generate = gr.Button(value="Generate").style(rounded=(False, True, True, False))
# image_out = gr.Image(height=512)
gallery = gr.Gallery(label="Generated images", show_label=False, elem_id="gallery").style(grid=[2], height="auto")
state_info = gr.Textbox(label="State", show_label=False, max_lines=2).style(container=False)
error_output = gr.Markdown()
with gr.Column(scale=45):
with gr.Tab("Options"):
with gr.Group():
neg_prompt = gr.Textbox(label="Negative prompt", placeholder="What to exclude from the image")
n_images = gr.Slider(label="Images", value=1, minimum=1, maximum=4, step=1)
with gr.Row():
guidance = gr.Slider(label="Guidance scale", value=7.5, maximum=15)
steps = gr.Slider(label="Steps", value=current_steps, minimum=2, maximum=75, step=1)
with gr.Row():
width = gr.Slider(label="Width", value=512, minimum=64, maximum=1024, step=8)
height = gr.Slider(label="Height", value=512, minimum=64, maximum=1024, step=8)
seed = gr.Slider(0, 2147483647, label='Seed (0 = random)', value=0, step=1)
with gr.Tab("Image to image"):
with gr.Group():
image = gr.Image(label="Image", height=256, tool="editor", type="pil")
strength = gr.Slider(label="Transformation strength", minimum=0, maximum=1, step=0.01, value=0.5)
if is_colab:
model_name.change(on_model_change, inputs=model_name, outputs=[custom_model_group, prompt], queue=False)
custom_model_path.change(custom_model_changed, inputs=custom_model_path, outputs=None)
# n_images.change(lambda n: gr.Gallery().style(grid=[2 if n > 1 else 1], height="auto"), inputs=n_images, outputs=gallery)
steps.change(on_steps_change, inputs=[steps], outputs=[], queue=False)
inputs = [model_name, prompt, guidance, steps, n_images, width, height, seed, image, strength, neg_prompt]
outputs = [gallery, error_output]
prompt.submit(inference, inputs=inputs, outputs=outputs)
generate.click(inference, inputs=inputs, outputs=outputs)
ex = gr.Examples([
[models[7].name, "tiny cute and adorable kitten adventurer dressed in a warm overcoat with survival gear on a winters day", 7.5, 25],
[models[4].name, "portrait of dwayne johnson", 7.0, 35],
[models[5].name, "portrait of a beautiful alyx vance half life", 10, 25],
[models[6].name, "Aloy from Horizon: Zero Dawn, half body portrait, smooth, detailed armor, beautiful face, illustration", 7.0, 30],
[models[5].name, "fantasy portrait painting, digital art", 4.0, 20],
], inputs=[model_name, prompt, guidance, steps], outputs=outputs, fn=inference, cache_examples=False)
gr.HTML("""
""")
demo.load(update_state_info, inputs=state_info, outputs=state_info, every=0.5, show_progress=False)
print(f"Space built in {time.time() - start_time:.2f} seconds")
# if not is_colab:
demo.queue(concurrency_count=1)
demo.launch(debug=is_colab, share=is_colab)
