import gradio as gr import requests import time import json import base64 import os from io import BytesIO import html import re import cv2 import torch from basicsr.archs.srvgg_arch import SRVGGNetCompact from gfpgan.utils import GFPGANer from realesrgan.utils import RealESRGANer os.system("pip freeze") # download weights if not os.path.exists('realesr-general-x4v3.pth'): os.system("wget https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.5.0/realesr-general-x4v3.pth -P .") if not os.path.exists('GFPGANv1.2.pth'): os.system("wget https://github.com/TencentARC/GFPGAN/releases/download/v1.3.0/GFPGANv1.2.pth -P .") if not os.path.exists('GFPGANv1.3.pth'): os.system("wget https://github.com/TencentARC/GFPGAN/releases/download/v1.3.0/GFPGANv1.3.pth -P .") if not os.path.exists('GFPGANv1.4.pth'): os.system("wget https://github.com/TencentARC/GFPGAN/releases/download/v1.3.0/GFPGANv1.4.pth -P .") if not os.path.exists('RestoreFormer.pth'): os.system("wget https://github.com/TencentARC/GFPGAN/releases/download/v1.3.4/RestoreFormer.pth -P .") if not os.path.exists('CodeFormer.pth'): os.system("wget https://github.com/TencentARC/GFPGAN/releases/download/v1.3.4/CodeFormer.pth -P .") # background enhancer with RealESRGAN model = SRVGGNetCompact(num_in_ch=3, num_out_ch=3, num_feat=64, num_conv=32, upscale=4, act_type='prelu') model_path = 'realesr-general-x4v3.pth' half = True if torch.cuda.is_available() else False upsampler = RealESRGANer(scale=4, model_path=model_path, model=model, tile=0, tile_pad=10, pre_pad=0, half=half) os.makedirs('output', exist_ok=True) class Prodia: def __init__(self, api_key, base=None): self.base = base or "https://api.prodia.com/v1" self.headers = { "X-Prodia-Key": api_key } def generate(self, params): response = self._post(f"{self.base}/sd/generate", params) return response.json() def transform(self, params): response = self._post(f"{self.base}/sd/transform", params) return response.json() def controlnet(self, params): response = self._post(f"{self.base}/sd/controlnet", params) return response.json() def get_job(self, job_id): response = self._get(f"{self.base}/job/{job_id}") return response.json() def wait(self, job): job_result = job while job_result['status'] not in ['succeeded', 'failed']: time.sleep(0.25) job_result = self.get_job(job['job']) return job_result def list_models(self): response = self._get(f"{self.base}/sd/models") return response.json() def list_samplers(self): response = self._get(f"{self.base}/sd/samplers") return response.json() def _post(self, url, params): headers = { **self.headers, "Content-Type": "application/json" } response = requests.post(url, headers=headers, data=json.dumps(params)) if response.status_code != 200: raise Exception(f"Bad Prodia Response: {response.status_code}") return response def _get(self, url): response = requests.get(url, headers=self.headers) if response.status_code != 200: raise Exception(f"Bad Prodia Response: {response.status_code}") return response def image_to_base64(image): # Convert the image to bytes buffered = BytesIO() image.save(buffered, format="PNG") # You can change format to PNG if needed # Encode the bytes to base64 img_str = base64.b64encode(buffered.getvalue()) return img_str.decode('utf-8') # Convert bytes to string def remove_id_and_ext(text): text = re.sub(r'\[.*\]$', '', text) extension = text[-12:].strip() if extension == "safetensors": text = text[:-13] elif extension == "ckpt": text = text[:-4] return text def get_data(text): results = {} patterns = { 'prompt': r'(.*)', 'negative_prompt': r'Negative prompt: (.*)', 'steps': r'Steps: (\d+),', 'seed': r'Seed: (\d+),', 'sampler': r'Sampler:\s*([^\s,]+(?:\s+[^\s,]+)*)', 'model': r'Model:\s*([^\s,]+)', 'cfg_scale': r'CFG scale:\s*([\d\.]+)', 'size': r'Size:\s*([0-9]+x[0-9]+)' } for key in ['prompt', 'negative_prompt', 'steps', 'seed', 'sampler', 'model', 'cfg_scale', 'size']: match = re.search(patterns[key], text) if match: results[key] = match.group(1) else: results[key] = None if results['size'] is not None: w, h = results['size'].split("x") results['w'] = w results['h'] = h else: results['w'] = None results['h'] = None return results def send_to_txt2img(image): result = {tabs: gr.update(selected="t2i")} try: text = image.info['parameters'] data = get_data(text) result[prompt] = gr.update(value=data['prompt']) result[negative_prompt] = gr.update(value=data['negative_prompt']) if data['negative_prompt'] is not None else gr.update() result[steps] = gr.update(value=int(data['steps'])) if data['steps'] is not None else gr.update() result[seed] = gr.update(value=int(data['seed'])) if data['seed'] is not None else gr.update() result[cfg_scale] = gr.update(value=float(data['cfg_scale'])) if data['cfg_scale'] is not None else gr.update() result[width] = gr.update(value=int(data['w'])) if data['w'] is not None else gr.update() result[height] = gr.update(value=int(data['h'])) if data['h'] is not None else gr.update() result[sampler] = gr.update(value=data['sampler']) if data['sampler'] is not None else gr.update() if model in model_names: result[model] = gr.update(value=model_names[model]) else: result[model] = gr.update() return result except Exception as e: print(e) return result prodia_client = Prodia(api_key=os.getenv("PRODIA_API_KEY")) model_list = prodia_client.list_models() model_names = {} for model_name in model_list: name_without_ext = remove_id_and_ext(model_name) model_names[name_without_ext] = model_name def txt2img(prompt, negative_prompt, model, steps, sampler, cfg_scale, width, height, seed): result = prodia_client.generate({ "prompt": prompt, "negative_prompt": negative_prompt, "model": model, "steps": steps, "sampler": sampler, "cfg_scale": cfg_scale, "width": width, "height": height, "seed": seed }) job = prodia_client.wait(result) return job["imageUrl"] def img2img(img, version, scale, weight): weight /= 100 print(img, version, scale, weight) try: extension = os.path.splitext(os.path.basename(str(img)))[1] img = cv2.imread(img, cv2.IMREAD_UNCHANGED) if len(img.shape) == 3 and img.shape[2] == 4: img_mode = 'RGBA' elif len(img.shape) == 2: # for gray inputs img_mode = None img = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR) else: img_mode = None if version == 'v1.2': face_enhancer = GFPGANer( model_path='GFPGANv1.2.pth', upscale=2, arch='clean', channel_multiplier=2, bg_upsampler=upsampler) elif version == 'v1.3': face_enhancer = GFPGANer( model_path='GFPGANv1.3.pth', upscale=2, arch='clean', channel_multiplier=2, bg_upsampler=upsampler) elif version == 'v1.4': face_enhancer = GFPGANer( model_path='GFPGANv1.4.pth', upscale=2, arch='clean', channel_multiplier=2, bg_upsampler=upsampler) elif version == 'RestoreFormer': face_enhancer = GFPGANer( model_path='RestoreFormer.pth', upscale=2, arch='RestoreFormer', channel_multiplier=2, bg_upsampler=upsampler) elif version == 'CodeFormer': face_enhancer = GFPGANer( model_path='CodeFormer.pth', upscale=2, arch='CodeFormer', channel_multiplier=2, bg_upsampler=upsampler) try: _, _, output = face_enhancer.enhance(img, has_aligned=False, only_center_face=False, paste_back=True, weight=weight) except RuntimeError as error: print('Error', error) try: interpolation = cv2.INTER_AREA if scale < 2 else cv2.INTER_LANCZOS4 h, w = img.shape[0:2] output = cv2.resize(output, (int(w * scale), int(h * scale)), interpolation=interpolation) except Exception as error: print('wrong scale input.', error) if img_mode == 'RGBA': # RGBA images should be saved in png format extension = 'png' else: extension = 'jpg' save_path = f'output/out.{extension}' cv2.imwrite(save_path, output) output = cv2.cvtColor(output, cv2.COLOR_BGR2RGB) return output except Exception as error: print('global exception', error) return None css = """ footer {visibility: hidden !important;} """ with gr.Blocks(css=css) as demo: with gr.Tabs() as tabs: with gr.Row(): with gr.Column(scale=3): with gr.Tab("Базовые настройки"): with gr.Row(): prompt = gr.Textbox(placeholder="Введите описание изображения...", show_label=True, label="Описание изображения", lines=3) with gr.Accordion(label="Модель нейросети:", open=False): model = gr.Radio(interactive=True, value="absolutereality_v181.safetensors [3d9d4d2b]", show_label=False, choices=prodia_client.list_models()) with gr.Tab("Расширенные настройки"): with gr.Row(): with gr.Row(): negative_prompt = gr.Textbox(placeholder="Negative Prompt", show_label=False, lines=3, value="[deformed | disfigured], poorly drawn, [bad : wrong] anatomy, [extra | missing | floating | disconnected] limb, (mutated hands and fingers), blurry") with gr.Column(scale=1): sampler = gr.Dropdown(value="DPM++ 2M Karras", show_label=True, label="Sampling Method", choices=prodia_client.list_samplers()) with gr.Column(scale=1): steps = gr.Slider(label="Sampling Steps", minimum=1, maximum=30, value=25, step=1) with gr.Column(scale=1): width = gr.Slider(label="Ширина", minimum=15, maximum=1024, value=512, step=8) height = gr.Slider(label="Длина", minimum=15, maximum=1024, value=512, step=8) cfg_scale = gr.Slider(label="CFG Scale", minimum=1, maximum=20, value=7, step=1) seed = gr.Slider(label="Seed", minimum=-1, maximum=10000000, value=-1) with gr.Tab("Настройки апскейлинга"): with gr.Row(): version = gr.Radio(choices=['v1.2', 'v1.3', 'v1.4', 'RestoreFormer', 'CodeFormer'], value='v1.4', label='Версия'), scale = gr.Number(label="Коэффициент масштабирования", value=2), weight = gr.Slider(0, 100, label='Weight, только для CodeFormer. 0 для лучшего качества, 100 для лучшей идентичности', value=50) with gr.Column(): text_button = gr.Button("Создать", variant='primary', elem_id="generate") with gr.Column(scale=2): image_output = gr.Image(show_label=True, label='Сгенерированное изображение:') with gr.Column(): text_button_up = gr.Button("Улучшить качество", variant='secondary', elem_id="upscb") with gr.Column(scale=2): image_output_up = gr.Image(show_label=True, label='Увеличенное изображение:') text_button.click(txt2img, inputs=[prompt, negative_prompt, model, steps, sampler, cfg_scale, width, height, seed], outputs=image_output) text_button_up.click(img2img, inputs=[image_output, version, scale, weight], outputs=image_output_up) demo.queue(concurrency_count=64, max_size=80, api_open=False).launch(max_threads=256)