from diffusers import StableDiffusionXLPipeline, StableDiffusionXLImg2ImgPipeline, EulerAncestralDiscreteScheduler import gradio as gr import torch from PIL import Image import random import os from huggingface_hub import hf_hub_download import torch from torch import autocast from diffusers import StableDiffusionXLPipeline, EulerAncestralDiscreteScheduler from safetensors import safe_open from compel import Compel, ReturnedEmbeddingsType from huggingface_hub import hf_hub_download model_id = 'aipicasso/emi' auth_token=os.environ["ACCESS_TOKEN"] scheduler = EulerAncestralDiscreteScheduler.from_pretrained(model_id, subfolder="scheduler", use_auth_token=auth_token) pipe = StableDiffusionXLPipeline.from_pretrained( model_id, torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32, scheduler=scheduler, use_auth_token=auth_token) #pipe.unet = torch.compile(pipe.unet, mode="reduce-overhead", fullgraph=True) #ckpt_file=hf_hub_download(repo_id=model_id, filename="v2.safetensors", token=auth_token) #pipe = StableDiffusionXLPipeline.from_single_file( # ckpt_file, # torch_dtype=torch.float16, # scheduler=scheduler #) #pipe.load_lora_weights("manual.safetensors") pipe_i2i = StableDiffusionXLImg2ImgPipeline.from_pretrained( model_id, torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32, scheduler=scheduler, use_auth_token=auth_token ) pipe=pipe.to("cuda") pipe_i2i=pipe_i2i.to("cuda") token_num=65 unaestheticXLv31="" embeddings_dict = {} with safe_open("unaestheticXLv31.safetensors", framework="pt") as f: for k in f.keys(): embeddings_dict[k] = f.get_tensor(k) pipe.text_encoder.resize_token_embeddings(len(pipe.tokenizer),pad_to_multiple_of=128) pipe.text_encoder_2.resize_token_embeddings(len(pipe.tokenizer),pad_to_multiple_of=128) for i in range(len(embeddings_dict["clip_l"])): token = f"sksd{chr(token_num)}" token_num+=1 unaestheticXLv31 += token pipe.tokenizer.add_tokens(token) token_id = pipe.tokenizer.convert_tokens_to_ids(token) pipe.text_encoder.get_input_embeddings().weight.data[token_id] = embeddings_dict["clip_l"][i] pipe.text_encoder_2.get_input_embeddings().weight.data[token_id] = embeddings_dict["clip_g"][i] unaestheticXLv1="" embeddings_dict = {} with safe_open("unaestheticXLv1.safetensors", framework="pt") as f: for k in f.keys(): embeddings_dict[k] = f.get_tensor(k) pipe.text_encoder.resize_token_embeddings(len(pipe.tokenizer),pad_to_multiple_of=128) pipe.text_encoder_2.resize_token_embeddings(len(pipe.tokenizer),pad_to_multiple_of=128) for i in range(len(embeddings_dict["clip_l"])): token = f"sksd{chr(token_num)}" token_num+=1 unaestheticXLv1 += token pipe.tokenizer.add_tokens(token) token_id = pipe.tokenizer.convert_tokens_to_ids(token) pipe.text_encoder.get_input_embeddings().weight.data[token_id] = embeddings_dict["clip_l"][i] pipe.text_encoder_2.get_input_embeddings().weight.data[token_id] = embeddings_dict["clip_g"][i] unaestheticXLv13="" embeddings_dict = {} with safe_open("unaestheticXLv13.safetensors", framework="pt") as f: for k in f.keys(): embeddings_dict[k] = f.get_tensor(k) pipe.text_encoder.resize_token_embeddings(len(pipe.tokenizer),pad_to_multiple_of=128) pipe.text_encoder_2.resize_token_embeddings(len(pipe.tokenizer),pad_to_multiple_of=128) for i in range(len(embeddings_dict["clip_l"])): token = f"sksd{chr(token_num)}" unaestheticXLv13 += token token_num+=1 pipe.tokenizer.add_tokens(token) token_id = pipe.tokenizer.convert_tokens_to_ids(token) pipe.text_encoder.get_input_embeddings().weight.data[token_id] = embeddings_dict["clip_l"][i] pipe.text_encoder_2.get_input_embeddings().weight.data[token_id] = embeddings_dict["clip_g"][i] compel = Compel(tokenizer=[pipe.tokenizer, pipe.tokenizer_2] , text_encoder=[pipe.text_encoder, pipe.text_encoder_2], returned_embeddings_type=ReturnedEmbeddingsType.PENULTIMATE_HIDDEN_STATES_NON_NORMALIZED, requires_pooled=[False, True]) def error_str(error, title="Error"): return f"""#### {title} {error}""" if error else "" def inference(prompt, guidance, steps, image_size="Landscape", seed=0, img=None, strength=0.5, neg_prompt="", disable_auto_prompt_correction=False, image_style="Animetic", original_model=False): generator = torch.Generator('cuda').manual_seed(seed) if seed != 0 else None prompt,neg_prompt=auto_prompt_correction(prompt,neg_prompt,disable_auto_prompt_correction) if(image_size=="Portrait"): height=1344 width=768 elif(image_size=="Landscape"): height=768 width=1344 else: height=1024 width=1024 print(prompt,neg_prompt) try: if img is not None: return img_to_img(prompt, neg_prompt, img, strength, guidance, steps, width, height, generator), None else: return txt_to_img(prompt, neg_prompt, guidance, steps, width, height, generator), None except Exception as e: return None, error_str(e) def auto_prompt_correction(prompt_ui,neg_prompt_ui,disable_auto_prompt_correction): # auto prompt correction prompt=str(prompt_ui) neg_prompt=str(neg_prompt_ui) prompt=prompt.lower() neg_prompt=neg_prompt.lower() if(disable_auto_prompt_correction): return prompt, neg_prompt if(prompt=="" and neg_prompt==""): prompt="1girl++, smile--, brown bob+++ hair, brown eyes, sunflowers, sky, transparent++" neg_prompt=f"({unaestheticXLv31})---, photo, deformed, realism, disfigured, low contrast, bad hand" return prompt, neg_prompt splited_prompt=prompt.replace(","," ").replace("_"," ").replace("+"," ").split(" ") human_words=["1girl","girl","maid","maids","female","1woman","woman","girls","2girls","3girls","4girls","5girls","a couple of girls","women","1boy","boy","boys","a couple of boys","2boys","male","1man","1handsome","1bishounen","man","men","guy","guys"] for word in human_words: if( word in splited_prompt): prompt=f"ime artwork, anime style, {prompt}" neg_prompt=f"({unaestheticXLv31})---,{neg_prompt}, photo, deformed, realism, disfigured, low contrast, bad hand" return prompt, neg_prompt animal_words=["cat","dog","bird","pigeon","rabbit","bunny","horse"] for word in animal_words: if( word in splited_prompt): prompt=f"anime style, a {prompt}, 4k, detailed" neg_prompt=f"{neg_prompt},({unaestheticXLv31})---" return prompt, neg_prompt background_words=["mount fuji","mt. fuji","building", "buildings", "tokyo", "kyoto", "nara", "shibuya", "shinjuku"] for word in background_words: if( word in splited_prompt): prompt=f"ime artwork, anime style, {prompt}, highly detailed" neg_prompt=f"girl, deformed+++, {neg_prompt}, girl, boy, photo, people, low quality, ui, error, lowres, jpeg artifacts, 2d, 3d, cg, text" return prompt, neg_prompt return prompt,neg_prompt def txt_to_img(prompt, neg_prompt, guidance, steps, width, height, generator): conditioning, pooled = compel([prompt, neg_prompt]) result = pipe( prompt_embeds=conditioning[0:1], pooled_prompt_embeds=pooled[0:1], negative_prompt_embeds=conditioning[1:2], negative_pooled_prompt_embeds=pooled[1:2], num_inference_steps = int(steps), guidance_scale = guidance, width = width, height = height, generator = generator) return result.images[0] def img_to_img(prompt, neg_prompt, img, strength, guidance, steps, width, height, generator): ratio = min(height / img.height, width / img.width) img = img.resize((int(img.width * ratio), int(img.height * ratio)), Image.LANCZOS) result = pipe_i2i( prompt, negative_prompt = neg_prompt, image = img, num_inference_steps = int(steps), strength = strength, guidance_scale = guidance, generator = generator) return result.images[0] css = """.main-div div{display:inline-flex;align-items:center;gap:.8rem;font-size:1.75rem}.main-div div h1{font-weight:900;margin-bottom:7px}.main-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=css) as demo: gr.HTML( f"""

Emi Demo

Demo for Emi

サンプル： そのままGenerateボタンを押してください。
sample : Click "Generate" button without any prompts.

sample prompt1 : 1girl++, cool+, smile--, colorful long hair, colorful eyes, stars, night, pastel color, transparent+

sample prompt2 : 1man+, focus, wavy short hair, blue eyes, black shirt, white background, simple background

sample prompt3 : anime style, 1girl++

共有ボタンを押してみんなに画像を共有しましょう。Please push share button to share your image.

Running on {"GPU 🔥" if torch.cuda.is_available() else f"CPU 🥶. For faster inference it is recommended to upgrade to GPU in Settings"}
Duplicate Space to say goodbye from waiting for the generating.

""" ) with gr.Row(): with gr.Column(scale=55): with gr.Group(): with gr.Row(): prompt = gr.Textbox(label="Prompt", show_label=False, max_lines=2,placeholder="[your prompt]") generate = gr.Button(value="Generate") image_out = gr.Image(height=768,width=1344) 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") disable_auto_prompt_correction = gr.Checkbox(label="Disable auto prompt corretion.") with gr.Row(): image_size=gr.Radio(["Portrait","Landscape","Square"]) image_size.show_label=False image_size.value="Square" with gr.Row(): guidance = gr.Slider(label="Guidance scale", value=7.5, maximum=25) steps = gr.Slider(label="Steps", value=20, minimum=2, maximum=75, step=1) 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) inputs = [prompt, guidance, steps, image_size, seed, image, strength, neg_prompt, disable_auto_prompt_correction] outputs = [image_out, error_output] prompt.submit(inference, inputs=inputs, outputs=outputs) generate.click(inference, inputs=inputs, outputs=outputs) demo.queue(concurrency_count=1) demo.launch()