File size: 7,493 Bytes
5b9202b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
import gradio as gr
import numpy as np
from PIL import Image
from transformers import pipeline
from diffusers import StableDiffusionDepth2ImgPipeline, StableDiffusionPipeline, StableDiffusionControlNetPipeline, StableDiffusionUpscalePipeline,  StableDiffusionImg2ImgPipeline, AutoPipelineForImage2Image
from diffusers import StableDiffusionControlNetPipeline, ControlNetModel, UniPCMultistepScheduler, DPMSolverMultistepScheduler

import torch
from controlnet_aux import HEDdetector
from diffusers.utils import load_image
from huggingface_hub import notebook_login, login
import concurrent.futures
from threading import Thread

hidden_booster_text = "beautiful face, beautiful hand, small boobs, a cup"
hidden_negative = "big boobs, huge boobs, sexy, dirty, d cup, e cup, g cup, slutty, badhandv4, ng_deepnegative_v1_75t, worst quality, low quality, extra digits, text, signature, bad anatomy, mutated hand, error, missing finger, cropped, worse quality, bad quality, lowres, floating limbs, bad hands, anatomical nonsense"

hed = HEDdetector.from_pretrained('lllyasviel/ControlNet')

controlnet_scribble = ControlNetModel.from_pretrained(
    "lllyasviel/sd-controlnet-scribble", torch_dtype=torch.float16
)

pipe_scribble = StableDiffusionControlNetPipeline.from_single_file(
    "https://huggingface.co/shellypeng/anime-god/blob/main/animeGod_v10.safetensors", controlnet=controlnet_scribble,
    torch_dtype=torch.float16,
)

pipe_scribble.load_textual_inversion("shellypeng/bad-prompt")
pipe_scribble.load_textual_inversion("shellypeng/badhandv4")
# pipe.load_textual_inversion("shellypeng/easynegative")
pipe_scribble.load_textual_inversion("shellypeng/deepnegative")
pipe_scribble.load_textual_inversion("shellypeng/verybadimagenegative")
pipe_scribble.scheduler = DPMSolverMultistepScheduler.from_config(pipe_scribble.scheduler.config, use_karras_sigmas=True)
# pipe.enable_model_cpu_offload()
def dummy(images, **kwargs):
    return images, False
pipe_scribble.safety_checker = lambda images, **kwargs: (images, [False] * len(images))
pipe_scribble.to("cuda")

def scribble_to_image(text, input_img):
    """
    pass the sd model and do scribble to image
    include Adetailer, detail tweaker lora, prompt backend include: beautiful eyes, beautiful face, beautiful hand, (maybe infer from user's prompt for gesture and facial
    expression to improve hand)
    """
    # change param "bag" below to text, image param below to input_img
    input_img = Image.fromarray(input_img)
    input_img = hed(input_img, scribble=True)
    input_img = load_image(input_img)
    # global prompt
    prompt = text + hidden_booster_text
    res_image0 = pipe_scribble(prompt, input_img, negative_prompt=hidden_negative, num_inference_steps=40).images[0]
    res_image1 = pipe_scribble(prompt, input_img, negative_prompt=hidden_negative, num_inference_steps=40).images[0]
    res_image2 = pipe_scribble(prompt, input_img, negative_prompt=hidden_negative, num_inference_steps=40).images[0]
    res_image3 = pipe_scribble(prompt, input_img, negative_prompt=hidden_negative, num_inference_steps=40).images[0]

    return res_image0, res_image1, res_image2, res_image3

theme = gr.themes.Soft(
    primary_hue="orange",
    secondary_hue="orange",
).set(
    block_background_fill='*primary_50'
)

pipe_img2img = StableDiffusionImg2ImgPipeline.from_single_file("https://huggingface.co/shellypeng/anime-god/blob/main/animeGod_v10.safetensors",
                                                               torch_dtype=torch.float16)
pipe_img2img.load_lora_weights("shellypeng/detail-tweaker")
pipe_img2img.fuse_lora(lora_scale=0.1)
pipe_img2img.load_textual_inversion("shellypeng/bad-prompt")
pipe_img2img.load_textual_inversion("shellypeng/badhandv4")
# pipe.load_textual_inversion("shellypeng/easynegative")
pipe_img2img.load_textual_inversion("shellypeng/deepnegative")
pipe_img2img.load_textual_inversion("shellypeng/verybadimagenegative")
pipe_img2img.scheduler = DPMSolverMultistepScheduler.from_config(pipe_img2img.scheduler.config, use_karras_sigmas=True)
# pipe.enable_model_cpu_offload()
def dummy(images, **kwargs):
    return images, False
pipe_img2img.safety_checker = lambda images, **kwargs: (images, [False] * len(images))
pipe_img2img = pipe_img2img.to("cuda")

def real_img2img_to_anime(text, input_img):
    """
    pass the sd model and do scribble to image
    include Adetailer, detail tweaker lora, prompt backend include: beautiful eyes, beautiful face, beautiful hand, (maybe infer from user's prompt for gesture and facial
    expression to improve hand)
    """
    input_img = Image.fromarray(input_img)
    input_img = load_image(input_img)
    prompt = text + hidden_booster_text
    # input_img = depth_estimator(input_img)['depth']
    res_image0 = pipe_img2img(prompt, input_img, negative_prompt=hidden_negative, num_inference_steps=40).images[0]
    res_image1 = pipe_img2img(prompt, input_img, negative_prompt=hidden_negative, num_inference_steps=40).images[0]
    res_image2 = pipe_img2img(prompt, input_img, negative_prompt=hidden_negative, num_inference_steps=40).images[0]
    res_image3 = pipe_img2img(prompt, input_img, negative_prompt=hidden_negative, num_inference_steps=40).images[0]

    return res_image0, res_image1, res_image2, res_image3


theme = gr.themes.Soft(
    primary_hue="orange",
    secondary_hue="orange",
).set(
    block_background_fill='*primary_50'
)


generator = torch.manual_seed(33)

with gr.Blocks(theme=theme, css="footer {visibility: hidden}", title="ShellAI Apps") as iface:
    with gr.Tab("Animefier"):
        with gr.Row(equal_height=True):
            with gr.Column():
                prompt_box = gr.Textbox(label="Prompt", placeholder="Enter a prompt")
                image_box = gr.Image(label="Input Image", height=350)
                gen_btn = gr.Button(value="Generate")
        with gr.Row(equal_height=True):
            global image1
            global image2
            global image3
            global image4
            image1 = gr.Image()
            image2 = gr.Image()
            image3 = gr.Image()
            image4 = gr.Image()

        def mult_thread(prompt_box, image_box):
            with concurrent.futures.ThreadPoolExecutor(max_workers=12000) as executor:
                future = executor.submit(real_img2img_to_anime, prompt_box, image_box)
                image1, image2, image3, image4 = future.result()
            return image1, image2, image3, image4

        gen_btn.click(mult_thread, [prompt_box, image_box], [image1, image2, image3, image4])
    with gr.Tab("AniSketch"):
        with gr.Row(equal_height=True):
            with gr.Column():
                prompt_box = gr.Textbox(label="Prompt", placeholder="Enter a prompt")
                image_box = gr.Image(label="Input Image", height=350)
                gen_btn = gr.Button(value="Generate")
        with gr.Row(equal_height=True):
            image1 = gr.Image()
            image2 = gr.Image()
            image3 = gr.Image()
            image4 = gr.Image()

        def mult_thread(prompt_box, image_box):
            with concurrent.futures.ThreadPoolExecutor(max_workers=12000) as executor:
                future = executor.submit(scribble_to_image, prompt_box, image_box)
                image1, image2, image3, image4 = future.result()
            return image1, image2, image3, image4

        gen_btn.click(mult_thread, [prompt_box, image_box], [image1, image2, image3, image4])

iface.launch(inline=False)