import gradio as gr import torch from diffusers import StableDiffusionImg2ImgPipeline from .utils.schedulers import SCHEDULER_LIST, get_scheduler_list from .utils.prompt2prompt import generate from .utils.device import get_device from PIL import Image from .download import get_share_js, get_community_loading_icon, CSS IMG2IMG_MODEL_LIST = { "OpenJourney v4" : "prompthero/openjourney-v4", "StableDiffusion 1.5" : "runwayml/stable-diffusion-v1-5", "StableDiffusion 2.1" : "stabilityai/stable-diffusion-2-1", "DreamLike 1.0" : "dreamlike-art/dreamlike-diffusion-1.0", "DreamLike 2.0" : "dreamlike-art/dreamlike-photoreal-2.0", "DreamShaper" : "Lykon/DreamShaper", "NeverEnding-Dream" : "Lykon/NeverEnding-Dream" } class StableDiffusionImage2ImageGenerator: def __init__(self): self.pipe = None def load_model(self, model_path, scheduler): model_path = IMG2IMG_MODEL_LIST[model_path] if self.pipe is None: self.pipe = StableDiffusionImg2ImgPipeline.from_pretrained( model_path, safety_checker=None, torch_dtype=torch.float32 ) device = get_device() self.pipe = get_scheduler_list(pipe=self.pipe, scheduler=scheduler) self.pipe.to(device) self.pipe.enable_attention_slicing() return self.pipe def generate_image( self, image_path: str, model_path: str, prompt: str, negative_prompt: str, scheduler: str, guidance_scale: int, num_inference_step: int, seed_generator=0, ): pipe = self.load_model( model_path=model_path, scheduler=scheduler, ) if seed_generator == 0: random_seed = torch.randint(0, 1000000, (1,)) generator = torch.manual_seed(random_seed) else: generator = torch.manual_seed(seed_generator) image = Image.open(image_path) images = pipe( prompt, image=image, negative_prompt=negative_prompt, num_images_per_prompt=1, num_inference_steps=num_inference_step, guidance_scale=guidance_scale, generator=generator, ).images return images def app(): demo = gr.Blocks(css=CSS) with demo: with gr.Row(): with gr.Column(): image2image_image_file = gr.Image( type="filepath", label="Upload",elem_id="image-upload-img2img" ).style(height=260) image2image_prompt = gr.Textbox( lines=1, placeholder="Prompt, keywords that describe the changes you want to apply to your image", show_label=False, elem_id="prompt-text-input-img2img", value='' ) image2image_negative_prompt = gr.Textbox( lines=1, placeholder="Negative Prompt, keywords that describe what you don't want in your image", show_label=False, elem_id = "negative-prompt-text-input-img2img", value='' ) # add button for generating a prompt from the prompt image2image_generate_prompt_button = gr.Button( label="Generate Prompt", type="primary", align="center", value = "Generate Prompt" ) # show a text box with the generated prompt image2image_generated_prompt = gr.Textbox( lines=1, placeholder="Generated Prompt", label = "Generated Prompt", show_label=True, info="Auto generated prompts for inspiration.", ) image2image_model_path = gr.Dropdown( choices=list(IMG2IMG_MODEL_LIST.keys()), value=list(IMG2IMG_MODEL_LIST.keys())[0], label="Imaget2Image Model Selection", elem_id="model-dropdown-img2img", info="Select the model you want to use for image2image generation." ) image2image_scheduler = gr.Dropdown( choices=SCHEDULER_LIST, value=SCHEDULER_LIST[0], label="Scheduler", elem_id="scheduler-dropdown-img2img", info="Scheduler list for models. Different schdulers result in different outputs." ) image2image_seed_generator = gr.Slider( label="Seed(0 for random)", minimum=0, maximum=1000000, value=0, elem_id="seed-slider-img2img", info="Set the seed to a specific value to reproduce the results." ) image2image_guidance_scale = gr.Slider( minimum=0.1, maximum=15, step=0.1, value=7.5, label="Guidance Scale", elem_id = "guidance-scale-slider-img2img", info = "Guidance scale determines how much the prompt will affect the image. Higher the value, more the effect." ) image2image_num_inference_step = gr.Slider( minimum=1, maximum=100, step=1, value=50, label="Num Inference Step", elem_id = "num-inference-step-slider-img2img", info = "Number of inference step determines the quality of the image. Higher the number, better the quality." ) image2image_predict_button = gr.Button(value="Generate image") with gr.Column(): output_image = gr.Gallery( label="Generated images", show_label=False, elem_id="gallery-img2img", ).style(grid=(1, 2)) with gr.Group(elem_id="container-advanced-btns"): with gr.Group(elem_id="share-btn-container"): community_icon_html, loading_icon_html = get_community_loading_icon("img2img") community_icon = gr.HTML(community_icon_html) loading_icon = gr.HTML(loading_icon_html) share_button = gr.Button("Save artwork", elem_id="share-btn-img2img") # Create an html for describing the models gr.HTML( """

Image2Image Models

Image2Image models are trained to generate images from a given prompt. The prompt should specify what to change in general for the provided image.

For example with a butterfly image, the prompt can be "blue butterfly".

Negative prompt can be used to specify what not to change in the image. For example, with the butterfly image, the negative prompt can be "dark blue, blurry image".

Stable Diffusion 1.5 & 2.1: Default model for many tasks.

OpenJourney v4: Generates fantasy themed images similar to the Midjourney model.

Dreamlike Photoreal 1.0 & 2.0 is SD 1.5 that generates realistic images.

""" ) image2image_predict_button.click( fn=StableDiffusionImage2ImageGenerator().generate_image, inputs=[ image2image_image_file, image2image_model_path, image2image_prompt, image2image_negative_prompt, image2image_scheduler, image2image_guidance_scale, image2image_num_inference_step, image2image_seed_generator, ], outputs=[output_image], ) image2image_generate_prompt_button.click( fn=generate, inputs=[image2image_prompt], outputs=[image2image_generated_prompt], ) return demo