Danbooru Prompt Selector/TEST2024/NAIA_generation (WEBUIA_0124).py

import requests
import random
from PIL import Image, ImageOps, ImageTk
from datetime import datetime
import time
from pathlib import Path
import io
import zipfile

#for webui
import base64
import re
import json

BASE_URL="https://api.novelai.net"

def make_turbo_prompt(gen_request):
    lines = gen_request['prompt']
    result = {
        "boys": False,
        "girls": False,
        "1girl": False,
        "1boy": False,
        "1other": False,
        "others": False
    }
    state = {
        "nude,": False,
        "pov,": False,
        "cum,": False,
        "after ": False,
        "pussy juice": False,
        "barefoot": False,
        "breasts": False,
        "ejaculation": False,
    }

    def insert_spaces(source_list, reference_list):
            modified_list = source_list.copy()
            for index, keyword in enumerate(reference_list):
                if keyword not in source_list:
                    space_count = len(keyword)  # 키워드 길이만큼의 공백 문자
                    modified_list.insert(index, ' ' * space_count)
            return modified_list

    keywords = gen_request['prompt'].split(', ')
    filtered_keywords = []
    removed_indices = []
    positive0, positive1, positive2, positive3 = gen_request.copy(),gen_request.copy(),gen_request.copy(),gen_request.copy()

    for word in result.keys():
        if word in lines:
            result[word] = True
    for word in state.keys():
        if word in gen_request['prompt']:
            state[word] = True

    key_index = int((len(keywords)/2)-1)

    if(result["1boy"]) or (result["boys"]):
            if(result["1girl"]):
                if('sex,' in gen_request['prompt']):
                    sex_pos_keywords  = ['stomach bulge','insertion', 'fucked silly', 'x-ray', 'orgasm', 'cross-section', 'uterus', 'overflow', 'rape', 'vaginal', 'anal']
                    facial_keywords = ['tongue','ahegao']
                    temp_sex_pos = []
                    temp_facial = []
                    cum_events = []
                    explicit_check = []
                    if 'open mouth' in keywords: keywords.remove('open mouth')
                    if 'closed mouth' in keywords: keywords.remove('closed mouth')
                    if 'after rape' in keywords: 
                        keywords.remove('after rape')
                        explicit_check.append('after rape')
                    for keyword in keywords:
                            if ('sex' not in keyword and 'cum' not in keyword and 'ejaculation' not in keyword and 'vaginal' not in keyword and 'penetration' not in keyword) and all(sex_pos not in keyword for sex_pos in sex_pos_keywords) and all(facial not in keyword for facial in facial_keywords):
                                filtered_keywords.append(keyword)
                            elif 'sex' in keyword:
                                removed_indices.append(keyword)
                            elif 'penetration' in keyword:
                                removed_indices.append(keyword)
                            elif 'cum' in keyword and keyword != 'cum':
                                cum_events.append(keyword)
                            elif any(sex_pos in keyword for sex_pos in sex_pos_keywords):
                                for sex_pos in sex_pos_keywords:
                                    if sex_pos in keyword:
                                        temp_sex_pos.append(sex_pos)
                            elif any(facial not in keyword for facial in facial_keywords):
                                for facial in facial_keywords:
                                    if facial in keyword:
                                        temp_facial.append(facial)
                    filtered_keywords.insert(int((len(filtered_keywords)/2)-1), ' no penetration, imminent penetration')
                    filtered_keywords_positive0 = filtered_keywords.copy()
                    filtered_keywords.remove(' no penetration, imminent penetration')
                    #0 imminent penetration, imminent sex
                    for i, keyword in enumerate(filtered_keywords):
                        if 'pantyhose' in keyword:
                            filtered_keywords[i] = 'torn ' + filtered_keywords[i]
                    #1 default
                    key_index = int((len(filtered_keywords)/2)-1)
                    if 'pussy' in filtered_keywords: key_index = filtered_keywords.index('pussy')
                    if 'penis' in filtered_keywords: key_index = filtered_keywords.index('penis')
                    filtered_keywords[key_index:key_index] = ['motion lines', 'surprised']
                    for keyword in removed_indices:
                        if 'cum' not in keyword and 'ejaculation' not in keyword:
                            filtered_keywords.insert(key_index,keyword)
                    if(temp_sex_pos): filtered_keywords[key_index:key_index] = temp_sex_pos
                    if('clothed sex' in filtered_keywords and not 'bottomless' in filtered_keywords): filtered_keywords.insert(filtered_keywords.index('clothed sex')+1, 'bottomless')
                    pos1_copied_keywords = filtered_keywords.copy()
                    for i, keyword in enumerate(pos1_copied_keywords):
                        if 'closed eyes' in keyword:
                            rand_num = random.randint(0,2)
                            if(rand_num == 0): pos1_copied_keywords[i] = 'half-' + pos1_copied_keywords[i]
                            elif(rand_num == 1 and 'closed eyes' in pos1_copied_keywords): 
                                pos1_copied_keywords.remove('closed eyes')
                                filtered_keywords[i] = 'half-closed eyes'
                    filtered_keywords_positive1 = pos1_copied_keywords.copy()
                    #2 ejaculation,cum in pussy
                    key_index = filtered_keywords.index('surprised')
                    filtered_keywords.remove('surprised')
                    filtered_keywords[key_index:key_index] = ["ejaculation","cum"]
                    for keyword in removed_indices:
                        if 'cum' in keyword:
                            filtered_keywords.insert(key_index,keyword)
                    if(temp_facial): filtered_keywords[key_index:key_index] =temp_facial
                    filtered_keywords_positive2 = filtered_keywords.copy()
                    #3 after sex, after ejaculation
                    for i, keyword in enumerate(filtered_keywords):
                        if 'closed eyes' in keyword:
                            rand_num = random.randint(0,2)
                            if(rand_num == 0 and  filtered_keywords[i] != 'half-closed eyes'): filtered_keywords[i] = 'half-' + filtered_keywords[i]
                            elif(rand_num == 1): filtered_keywords[i] = 'empty eyes'
                            else: filtered_keywords[i] = 'empty eyes, half-closed eyes'
                    if 'sex' in filtered_keywords:
                        key_index = filtered_keywords.index('sex')
                    elif 'group sex' in filtered_keywords:
                        key_index = filtered_keywords.index('group sex')
                    filtered_keywords.remove('ejaculation')
                    filtered_keywords[key_index:key_index] = ['cum drip', 'erection'] + cum_events
                    if(explicit_check): filtered_keywords[key_index:key_index] = explicit_check
                    if 'sex' in filtered_keywords and 'group sex' not in filtered_keywords:
                        if('pussy' in filtered_keywords and not 'anal' in filtered_keywords): filtered_keywords.insert(filtered_keywords.index('sex')+1, 'after vaginal, spread pussy')
                        elif('anal' in filtered_keywords): filtered_keywords.insert(filtered_keywords.index('sex')+1, 'after anus, cum in ass')
                        filtered_keywords.insert(filtered_keywords.index('sex'), 'after sex')
                        filtered_keywords.remove('sex')
                    elif 'group sex' in filtered_keywords:
                        if('vaginal' in filtered_keywords and not 'anal' in filtered_keywords): 
                            filtered_keywords.insert(filtered_keywords.index('group sex')+1, 'after vaginal, spread pussy')
                            if 'multiple penises' in filtered_keywords: filtered_keywords.insert(filtered_keywords.index('group sex')+3, 'cum on body, bukkake')
                        elif('anal' in filtered_keywords): 
                            filtered_keywords.insert(filtered_keywords.index('group sex')+1, 'after anus, cum in ass')
                            if 'multiple penises' in filtered_keywords: filtered_keywords.insert(filtered_keywords.index('group sex')+3, 'cum on body, bukkake')
                        else: filtered_keywords.insert(filtered_keywords.index('group sex')+1, 'cum on body, {bukkake}')
                    temp_post_keyword = []
                    for keyword in sex_pos_keywords:
                        if not (keyword == 'orgasm' or keyword == 'overflow'):
                            if keyword in filtered_keywords:
                                temp_post_keyword.append(keyword)
                    for keyword in temp_post_keyword:
                        filtered_keywords.remove(keyword)

                    positive0['prompt'] = ', '.join(insert_spaces(filtered_keywords_positive0, filtered_keywords)).strip()
                    positive1['prompt'] = ', '.join(insert_spaces(filtered_keywords_positive1, filtered_keywords)).strip()
                    positive2['prompt'] = ', '.join(insert_spaces(filtered_keywords_positive2, filtered_keywords)).strip()
                    positive3['prompt'] = ', '.join(filtered_keywords).strip()
                    positive0["type"] = "turbo"
                    positive1["type"] = "turbo"
                    positive2["type"] = "turbo"
                    positive3["type"] = "turbo"
    return positive0, positive1, positive2, positive3

def generate_image_NAI(access_token, prompt, model, action, parameters):
    data = {
        "input": prompt,
        "model": model,
        "action": action,
        "parameters": parameters,
    }

    response = requests.post(f"{BASE_URL}/ai/generate-image", json=data, headers={ "Authorization": f"Bearer {access_token}" })
    # catch any errors
    return response.content

# WebUI
def convert_prompt(prompt):
    keywords = [keyword.strip() for keyword in prompt.split(',')]
    converted_keywords = []
    for keyword in keywords:
        if '{' in keyword:
            keyword = keyword.replace('{', '(').replace('}', ')')
        converted_keywords.append(keyword)
    return ', '.join(converted_keywords)

def generate_image_webui(access_token, prompt, model, action, parameters):
    samplers = {
        "k_euler": "Euler",
        "k_euler_ancestral": "Euler a",
        "k_dpmpp_2s_ancestral": "DPM++ 2S a",
        "k_dpmpp_sde": "DPM++ SDE"
    }

    # matching data format
    data = {
        "input": prompt,
        "model": model,
        "action": action,
        "parameters": parameters,
    }

    params = {
        "prompt": convert_prompt(data['input']),
        "negative_prompt": convert_prompt(data['parameters']['negative_prompt']),
        "steps": data['parameters']['steps'],
        "width": data['parameters']['width'],
        "height": data['parameters']['height'],
        "cfg_scale": data['parameters']['scale'],
        "sampler_index": samplers[data['parameters']['sampler']],
        "seed": data['parameters']['seed'],
        "seed_resize_from_h": -1,
        "seed_resize_from_w": -1,
        "denoising_strength": None,
        "n_iter": "1",
        "batch_size": data['parameters']['n_samples']
    }

    if data['parameters']['enable_hr'] == True:
        params['enable_hr'] = True
        params["hr_upscaler"] = data['parameters']["hr_upscaler"]
        params["hr_scale"] = data['parameters']["hr_scale"]
        params["hr_second_pass_steps"] = data['parameters']["hr_second_pass_steps"]
        params["denoising_strength"] = data['parameters']["denoising_strength"]

    if data['parameters']['enable_AD'] == True:
        params["alwayson_scripts"] = {"ADetailer": 
            {
                "args": [
                    True,
                    False,
                    {
                        "ad_model": "face_yolov8n.pt",
                        "ad_prompt": params["prompt"],
                        "ad_negative_prompt": params["negative_prompt"],
                        "ad_inpaint_only_masked": True,
                        "ad_inpaint_only_masked_padding": 32,
                        "ad_sampler" : samplers[data['parameters']['sampler']]
                    }
                ]
            }
        }

    res = requests.post(f"{access_token}/sdapi/v1/txt2img", json=params)
    imageb64s = res.json()['images']
    content = None
    for b64 in imageb64s:
        img = b64.encode()
        content = base64.b64decode(img)

    s = io.BytesIO()
    zf = zipfile.ZipFile(s, "w")
    zf.writestr("generated", content)
    zf.close()
    return s.getvalue()

# WebUI
def generate_image(access_token, prompt, model, action, parameters):
    if re.match(r'^http[s]?://', access_token):
        return generate_image_webui(**locals())
    return generate_image_NAI(**locals())

def generate(gen_request):

    params = {
        "legacy": False,
        "quality_toggle": True if gen_request["quality_toggle"] == 1 else False,
        "width": gen_request["width"],
        "height": gen_request["height"],
        "n_samples": 1,
        "seed": gen_request["seed"],
        "extra_noise_seed": random.randint(0,9999999999),
        "sampler": gen_request["sampler"],
        "steps": gen_request["steps"],
        "scale": gen_request["scale"],
        "uncond_scale": 1.0,
        "negative_prompt": gen_request["negative"],
        "sm" : gen_request["sema"],
        "sm_dyn" : gen_request["sema_dyn"],
        "decrisper": False,
        "controlnet_strength": 1.0,
        "add_original_image": False,
        "cfg_rescale": gen_request["cfg_rescale"],
        "noise_schedule": "native",
        "enable_hr" : gen_request["enable_hr"],
        "enable_AD" : gen_request["enable_AD"]
        }
    
    if params["enable_hr"] == True:
        params["hr_upscaler"] = gen_request["hr_upscaler"]
        params["hr_scale"] = gen_request["hr_scale"]
        params["hr_second_pass_steps"] = gen_request["hr_second_pass_steps"]
        params["denoising_strength"] = gen_request["denoising_strength"]
    
    # 와일드카드 기능 만들어야함
    positive = gen_request["prompt"]

    filename_rule = gen_request["png_rule"]
    save_folder = gen_request["save_folder"]

    access_token = gen_request["access_token"]
    additional_folder = ""

    def resize_and_fill(image, max_size=None):
        if max_size is None:
            max_size = gen_request["user_screen_size"]
        original_width, original_height = image.size
        if original_width > max_size or original_height > max_size:
            # 비율을 유지하면서 크기 조정
            image.thumbnail((max_size, max_size))

            # 새 이미지 크기 계산
            width, height = image.size
            new_image = Image.new("RGB", (max_size, max_size), "black")
            new_image.paste(image, ((max_size - width) // 2, (max_size - height) // 2))
            return new_image
        else:
            return image

    def log_error(e, output_file_path="output_file_path"):
        # 현재 시간을 얻습니다
        current_time = datetime.now().strftime("%m/%d %H:%M:%S")

        # 에러 로그 메시지
        error_message = f"#### Error occured at {current_time} ####\nError: {e}\n############################################\n"

        # 지정된 출력 폴더의 error_log.txt 파일에 쓰기
        with open(f"error_log.txt", "a") as file:
            file.write(error_message)

    try:
        zipped_bytes = generate_image(access_token, positive, "nai-diffusion-3", "generate", params)
        if gen_request["png_rule"] == "count":
            additional_folder = "/" + gen_request["start_time"]
        if gen_request["type"] == "turbo":
            additional_folder += "/turbo"
        d = Path(save_folder + additional_folder)
        d.mkdir(parents=True, exist_ok=True)
        zipped = zipfile.ZipFile(io.BytesIO(zipped_bytes))
        image_bytes = zipped.read(zipped.infolist()[0])
        if gen_request["png_rule"] == "count": 
            _count = gen_request["count"]
            filename = (d / f"{_count:05}.png" )
        else: filename = (d / f"{datetime.now().strftime('%Y%m%d_%H%M%S')}.png" )
        filename.write_bytes(image_bytes)
        i = Image.open(io.BytesIO(image_bytes))
        i = ImageOps.exif_transpose(i).convert("RGB")
        i_resized = resize_and_fill(i)
        #tk_image = ImageTk.PhotoImage(i_resized)
        return i_resized, positive, params['seed'], i.info, str(filename)
    except Exception as e:
        try:
            if zipped_bytes is None:
                raise ValueError("Connection broken (Protocol Error)")
            error_message = zipped_bytes.decode('utf-8')[2:-2]
        except Exception as inner_exception:
            error_message = str(inner_exception)
        log_error(error_message, "path_to_output_folder")
        return None, error_message, params['seed'], None, None