app.py

import gradio as gr
import numpy as np
import random
import spaces  # [uncomment to use ZeroGPU]
from diffusers import DiffusionPipeline
import torch
import io
from PIL import Image
from tags import participant_tags, tribe_tags, skin_tone_tags, body_type_tags, tattoo_tags, piercing_tags, expression_tags, eye_tags, hair_style_tags, position_tags, fetish_tags, location_tags, camera_tags, atmosphere_tags

device = "cuda" if torch.cuda.is_available() else "cpu"
model_repo_id = "John6666/wai-ani-nsfw-ponyxl-v8-sdxl"  # Replace with your desired model

if torch.cuda.is_available():
    torch_dtype = torch.float16
else:
    torch_dtype = torch.float32

pipe = DiffusionPipeline.from_pretrained(model_repo_id, torch_dtype=torch_dtype)
pipe = pipe.to(device)

MAX_SEED = np.iinfo(np.int32).max
MAX_IMAGE_SIZE = 2048

@spaces.GPU  # [uncomment to use ZeroGPU]
def infer(prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps,
          selected_participant_tags, selected_tribe_tags, selected_skin_tone_tags, selected_body_type_tags,
          selected_tattoo_tags, selected_piercing_tags, selected_expression_tags, selected_eye_tags,
          selected_hair_style_tags, selected_position_tags, selected_fetish_tags, selected_location_tags,
          selected_camera_tags, selected_atmosphere_tags, active_tab, progress=gr.Progress(track_tqdm=True)):

    if active_tab == "Prompt Input":
        final_prompt = f'score_9, score_8_up, score_7_up, source_anime, {prompt}'
    else:
        selected_tags = (
            [participant_tags[tag] for tag in selected_participant_tags] +
            [tribe_tags[tag] for tag in selected_tribe_tags] +
            [skin_tone_tags[tag] for tag in selected_skin_tone_tags] +
            [body_type_tags[tag] for tag in selected_body_type_tags] +
            [tattoo_tags[tag] for tag in selected_tattoo_tags] +
            [piercing_tags[tag] for tag in selected_piercing_tags] +
            [expression_tags[tag] for tag in selected_expression_tags] +
            [eye_tags[tag] for tag in selected_eye_tags] +
            [hair_style_tags[tag] for tag in selected_hair_style_tags] +
            [position_tags[tag] for tag in selected_position_tags] +
            [fetish_tags[tag] for tag in selected_fetish_tags] +
            [location_tags[tag] for tag in selected_location_tags] +
            [camera_tags[tag] for tag in selected_camera_tags] +
            [atmosphere_tags[tag] for tag in selected_atmosphere_tags]
        )
        tags_text = ', '.join(selected_tags)
        final_prompt = f'score_9, score_8_up, score_7_up, source_anime, {tags_text}'

    additional_negatives = "worst quality, bad quality, jpeg artifacts, source_cartoon, 3d, (censor), monochrome, blurry, lowres, watermark"
    full_negative_prompt = f"{additional_negatives}, {negative_prompt}"

    if randomize_seed:
        seed = random.randint(0, MAX_SEED)

    generator = torch.Generator().manual_seed(seed)

    # Generate the image
    image = pipe(
        prompt=final_prompt,
        negative_prompt=full_negative_prompt,
        guidance_scale=guidance_scale,
        num_inference_steps=num_inference_steps,
        width=width,
        height=height,
        generator=generator
    ).images[0]

    # Convert the image to the desired format (JPEG or PNG)
    img_byte_arr = io.BytesIO()
    image_format = "PNG"  # You can change this to "JPEG" if you prefer JPEG
    image.save(img_byte_arr, format=image_format)
    img_byte_arr.seek(0)  # Reset pointer to the beginning of the BytesIO object

    # Return the image as a file (PNG or JPEG), seed, and the used prompts
    return img_byte_arr, seed, f"Prompt used: {final_prompt}\nNegative prompt used: {full_negative_prompt}"


css = """
#col-container {
    margin: 0 auto;
    max-width: 1280px;
}

#left-column {
    width: 50%;
    display: inline-block;
    padding-right: 20px;
    padding-left: 20px;
    vertical-align: top;
}

#right-column {
    width: 50%;
    display: inline-block;
    vertical-align: top;
    padding-left: 20px;
    margin-top: 53px;
}

#left-column > * {
    margin-bottom: 20px;
}

#run-button {
    width: 100%;
    margin-top: 10px;
    display: block;
}

#prompt-info {
    margin-bottom: 20px;
}

#result {
    margin-bottom: 20px;
}

.gradio-tabs > .tab-item {
    margin-bottom: 20px;
}

#prompt {
    margin-bottom: 20px;
}
"""

with gr.Blocks(css=css) as demo:

    with gr.Row():
        with gr.Column(elem_id="left-column"):
            gr.Markdown("""# Rainbow Media X""")

            # Display result image at the top
            result = gr.Image(label="Result", show_label=False, elem_id="result")

            # Add a textbox to display the prompts used for generation
            prompt_info = gr.Textbox(label="Prompts Used", lines=3, interactive=False, elem_id="prompt-info")

            # Advanced Settings and Run Button
            with gr.Accordion("Advanced Settings", open=False):
                negative_prompt = gr.Textbox(
                    label="Negative prompt",
                    max_lines=1,
                    placeholder="Enter a negative prompt",
                    visible=True,
                )

                seed = gr.Slider(
                    label="Seed",
                    minimum=0,
                    maximum=MAX_SEED,
                    step=1,
                    value=0,
                )

                randomize_seed = gr.Checkbox(label="Randomize seed", value=True)

                with gr.Row():
                    width = gr.Slider(
                        label="Width",
                        minimum=256,
                        maximum=MAX_IMAGE_SIZE,
                        step=32,
                        value=1024,
                    )

                    height = gr.Slider(
                        label="Height",
                        minimum=256,
                        maximum=MAX_IMAGE_SIZE,
                        step=32,
                        value=1024,
                    )

                with gr.Row():
                    guidance_scale = gr.Slider(
                        label="Guidance scale",
                        minimum=0.0,
                        maximum=10.0,
                        step=0.1,
                        value=7,
                    )

                    num_inference_steps = gr.Slider(
                        label="Number of inference steps",
                        minimum=1,
                        maximum=50,
                        step=1,
                        value=35,
                    )

            # Full-width "Run" button
            run_button = gr.Button("Run", elem_id="run-button")

        with gr.Column(elem_id="right-column"):
            # Removed the Prompt / Tag Input title here
            # State to track active tab
            active_tab = gr.State("Prompt Input")

            # Tabbed interface to select either Prompt or Tags
            with gr.Tabs() as tabs:
                with gr.TabItem("Prompt Input") as prompt_tab:
                    prompt = gr.Textbox(
                        label="Prompt",
                        show_label=False,
                        max_lines=1,
                        placeholder="Enter your prompt",
                        container=False,
                        elem_id="prompt"
                    )
                    prompt_tab.select(lambda: "Prompt Input", inputs=None, outputs=active_tab)

                with gr.TabItem("Tag Selection") as tag_tab:
                    # Tag selection checkboxes for each tag group
                    selected_participant_tags = gr.CheckboxGroup(choices=list(participant_tags.keys()), label="Participant Tags")
                    selected_tribe_tags = gr.CheckboxGroup(choices=list(tribe_tags.keys()), label="Tribe Tags")
                    selected_skin_tone_tags = gr.CheckboxGroup(choices=list(skin_tone_tags.keys()), label="Skin Tone Tags")
                    selected_body_type_tags = gr.CheckboxGroup(choices=list(body_type_tags.keys()), label="Body Type Tags")
                    selected_tattoo_tags = gr.CheckboxGroup(choices=list(tattoo_tags.keys()), label="Tattoo Tags")
                    selected_piercing_tags = gr.CheckboxGroup(choices=list(piercing_tags.keys()), label="Piercing Tags")
                    selected_expression_tags = gr.CheckboxGroup(choices=list(expression_tags.keys()), label="Expression Tags")
                    selected_eye_tags = gr.CheckboxGroup(choices=list(eye_tags.keys()), label="Eye Tags")
                    selected_hair_style_tags = gr.CheckboxGroup(choices=list(hair_style_tags.keys()), label="Hair Style Tags")
                    selected_position_tags = gr.CheckboxGroup(choices=list(position_tags.keys()), label="Position Tags")
                    selected_fetish_tags = gr.CheckboxGroup(choices=list(fetish_tags.keys()), label="Fetish Tags")
                    selected_location_tags = gr.CheckboxGroup(choices=list(location_tags.keys()), label="Location Tags")
                    selected_camera_tags = gr.CheckboxGroup(choices=list(camera_tags.keys()), label="Camera Tags")
                    selected_atmosphere_tags = gr.CheckboxGroup(choices=list(atmosphere_tags.keys()), label="Atmosphere Tags")
                    tag_tab.select(lambda: "Tag Selection", inputs=None, outputs=active_tab)

        run_button.click(
            infer,
            inputs=[prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps,
                    selected_participant_tags, selected_tribe_tags, selected_skin_tone_tags, selected_body_type_tags,
                    selected_tattoo_tags, selected_piercing_tags, selected_expression_tags, selected_eye_tags,
                    selected_hair_style_tags, selected_position_tags, selected_fetish_tags, selected_location_tags,
                    selected_camera_tags, selected_atmosphere_tags, active_tab],
            outputs=[result, seed, prompt_info]
        )

demo.queue().launch()