import spaces import os import gradio as gr import torch import numpy as np import random from diffusers import FluxPipeline from translatepy import Translator from huggingface_hub import hf_hub_download import requests import re os.environ["HF_HUB_ENABLE_HF_TRANSFER"] = "1" translator = Translator() HF_TOKEN = os.environ.get("HF_TOKEN", None) # Constants model = "black-forest-labs/FLUX.1-dev" default_lora = "Shakker-Labs/FilmPortrait" default_weight_name = 'filmfotos.safetensors' MAX_SEED = np.iinfo(np.int32).max CSS = """ footer { visibility: hidden; } """ JS = """function () { gradioURL = window.location.href if (!gradioURL.endsWith('?__theme=dark')) { window.location.replace(gradioURL + '?__theme=dark'); } }""" if torch.cuda.is_available(): pipe = FluxPipeline.from_pretrained(model, torch_dtype=torch.bfloat16) pipe.load_lora_weights(default_lora, weight_name = default_weight_name) # default load lora pipe.fuse_lora(lora_scale=0.9) def scrape_lora_link(url): try: # Send a GET request to the URL response = requests.get(url) response.raise_for_status() # Raise an exception for bad status codes # Get the content of the page content = response.text # Use regular expression to find the link pattern = r'href="(.*?lora.*?\.safetensors\?download=true)"' match = re.search(pattern, content) if match: safetensors_url = match.group(1) filename = safetensors_url.split('/')[-1].split('?')[0] # Extract the filename from the URL return filename else: return None except requests.RequestException as e: raise gr.Error(f"An error occurred while fetching the URL: {e}") def enable_lora(lora_scale,lora_add,progress=gr.Progress(track_tqdm=True)): pipe.unload_lora_weights() if not lora_add: gr.Info("No Lora Loaded, Use basemodel") return gr.update(value="") else: url = f'https://huggingface.co/{lora_add}/tree/main' lora_name = scrape_lora_link(url) if lora_name: print(f'lora loading: {lora_add}/{lora_name}') pipe.load_lora_weights(lora_add, weight_name=lora_name) pipe.fuse_lora(lora_scale=lora_scale) gr.Info(f"{lora_add} Loaded") return gr.update(label="LoRA Loaded Now") else: try: pipe.load_lora_weights(lora_add) pipe.fuse_lora(lora_scale=lora_scale) gr.Info(f"{lora_add} Loaded") return gr.update(label="LoRA Loaded Now") except: raise gr.Error(f"{lora_add} Load fail, check again.") @spaces.GPU() def generate_image( prompt:str, lora_word:str, width:int=768, height:int=1024, scales:float=3.5, steps:int=24, seed:int=-1, nums:int=1, progress=gr.Progress(track_tqdm=True)): pipe.to("cuda") if seed == -1: seed = random.randint(0, MAX_SEED) seed = int(seed) text = str(translator.translate(prompt, 'English')) + "," + lora_word print(f"Prompt: {text}") generator = torch.Generator().manual_seed(seed) image = pipe( prompt=text, height=height, width=width, guidance_scale=scales, output_type="pil", num_inference_steps=steps, max_sequence_length=512, num_images_per_prompt=nums, generator=generator, ).images return image, seed examples = [ ["close up portrait, Amidst the interplay of light and shadows in a photography studio,a soft spotlight traces the contours of a face,highlighting a figure clad in a sleek black turtleneck. The garment,hugging the skin with subtle luxury,complements the Caucasian model's understated makeup,embodying minimalist elegance. Behind,a pale gray backdrop extends,its fine texture shimmering subtly in the dim light,artfully balancing the composition and focusing attention on the subject. In a palette of black,gray,and skin tones,simplicity intertwines with profundity,as every detail whispers untold stories.",0.9,"Shakker-Labs/AWPortrait-FL",""], ["Caucasian,The image features a young woman of European descent standing in an studio setting,surrounded by silk. (She is wearing a silk dress),paired with a bold. Her brown hair is wet and tousled,falling naturally around her face,giving her a raw and edgy look. The woman has an intense and direct gaze,adding to the dramatic feel of the image. The backdrop is flowing silk,big silk. The overall composition blends elements of fashion and nature,creating a striking and powerful visual",0.9,"Shakker-Labs/AWPortrait-FL",""], ["A young Japanese girl, profile, blue hours, Tokyo tower",0.9,"Shakker-Labs/FilmPortrait","filmfotos, film grain, reversal film photography"], ["A young asian girl",0.9,"Shakker-Labs/FilmPortrait","filmfotos, film grain, reversal film photography"] ] # Gradio Interface with gr.Blocks(css=CSS, js=JS, theme="Nymbo/Nymbo_Theme") as demo: gr.HTML("

Flux Labs

") gr.HTML("

Load the LoRA model on the menu

") with gr.Row(): with gr.Column(scale=4): img = gr.Gallery(label='flux Generated Image', columns = 1, preview=True, height=600) with gr.Row(): prompt = gr.Textbox(label='Enter Your Prompt (Multi-Languages)', lines=2, placeholder="Enter prompt...", scale=6) sendBtn = gr.Button(scale=1, variant='primary') with gr.Accordion("Advanced Options", open=True): with gr.Column(scale=1): width = gr.Slider( label="Width", minimum=512, maximum=1280, step=8, value=768, ) height = gr.Slider( label="Height", minimum=512, maximum=1280, step=8, value=1024, ) scales = gr.Slider( label="Guidance", minimum=3.5, maximum=7, step=0.1, value=3.5, ) steps = gr.Slider( label="Steps", minimum=1, maximum=100, step=1, value=24, ) seed = gr.Slider( label="Seeds", minimum=-1, maximum=MAX_SEED, step=1, value=-1, ) nums = gr.Slider( label="Image Numbers", minimum=1, maximum=4, step=1, value=1, ) with gr.Column(scale=1): lora_scale = gr.Slider( label="LoRA Scale", minimum=0.1, maximum=1.0, step=0.1, value=0.9, ) lora_add = gr.Textbox( label="Flux LoRA", info="Copy the HF LoRA model name here", lines=1, value="Shakker-Labs/FilmPortrait", ) lora_word = gr.Textbox( label="Add Flux LoRA Trigger Word", info="Add the Trigger Word", lines=1, value="filmfotos, film grain, reversal film photography", ) load_lora = gr.Button(value="Load LoRA", variant='secondary') gr.Examples( examples=examples, inputs=[prompt,lora_scale,lora_add,lora_word], cache_examples=False, examples_per_page=4, ) load_lora.click(fn=enable_lora, inputs=[lora_scale,lora_add], outputs=lora_add) gr.on( triggers=[ prompt.submit, sendBtn.click, ], fn=generate_image, inputs=[ prompt, lora_word, width, height, scales, steps, seed, nums ], outputs=[img, seed], api_name="run", ) demo.queue().launch()