app.py

import os
import gradio as gr
import numpy as np
import random
from huggingface_hub import AsyncInferenceClient
from translatepy import Translator
import requests
import re
import asyncio
from PIL import Image
from gradio_client import Client, handle_file
from huggingface_hub import login
from gradio_imageslider import ImageSlider


translator = Translator()
HF_TOKEN = os.environ.get("HF_TOKEN", None)
basemodel = "black-forest-labs/FLUX.1-schnell"
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'); } }"


def enable_lora(lora_add): 
    return basemodel if not lora_add else lora_add


async def generate_image(prompt, model, lora_word, width, height, scales, steps, seed):
    if seed == -1: 
        seed = random.randint(0, MAX_SEED)
    seed = int(seed)
    text = str(translator.translate(prompt, 'English')) + "," + lora_word
    client = AsyncInferenceClient()
    try: 
        image = await client.text_to_image(prompt=text, height=height, width=width, guidance_scale=scales, num_inference_steps=steps, model=model)
    except Exception as e: 
        raise gr.Error(f"Error in {e}")
    return image, seed


async def gen(prompt, lora_add, lora_word, width, height, scales, steps, seed, upscale_factor, process_upscale):
    model = enable_lora(lora_add)
    image, seed = await generate_image(prompt, model, lora_word, width, height, scales, steps, seed)
    image_path = "temp_image.png"
    image.save(image_path)
    
    if process_upscale:
        upscale_image = get_upscale_finegrain(prompt, image_path, upscale_factor)
    else:
        upscale_image = image_path
    
    return [image_path, upscale_image]


def get_upscale_finegrain(prompt, img_path, upscale_factor):
    client = Client("finegrain/finegrain-image-enhancer")
    result = client.predict(input_image=handle_file(img_path), prompt=prompt, negative_prompt="", seed=42, upscale_factor=upscale_factor, controlnet_scale=0.6, controlnet_decay=1, condition_scale=6, tile_width=112, tile_height=144, denoise_strength=0.35, num_inference_steps=18, solver="DDIM", api_name="/process")
    return result[1]


css = """
#col-container{
    margin: 0 auto;
    max-width: 1024px;
}
"""


with gr.Blocks(css=css) as demo:
    with gr.Column(elem_id="col-container"):
        gr.Markdown("# Flux Upscaled")
        gr.Markdown("Step 1: Generate image with FLUX schnell; Step 2: UpScale with Finegrain Image-Enhancer")
        with gr.Group():
            prompt = gr.Textbox(label="Prompt")
            with gr.Row():
                lora_add = gr.Textbox(label="Add Flux LoRA", info="Copy the HF LoRA model name here", lines=1, placeholder="Please use Warm status model")
                lora_word = gr.Textbox(label="Add Flux LoRA Trigger Word", info="Add the Trigger Word", lines=1, value="")
                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)
                upscale_factor = gr.Radio(label="UpScale Factor", choices=[2, 3, 4], value=2, scale=2)
                process_upscale = gr.Checkbox(label="Process Upscale", value=True)
                submit_btn = gr.Button("Submit", scale=1)
            output_res = ImageSlider(label="Flux / Upscaled")

        submit_btn.click(
            fn=lambda: None,
            inputs=None,
            outputs=[output_res],
            queue=False
        ).then(
            fn=gen,
            inputs=[prompt, lora_add, lora_word, width, height, scales, steps, seed, upscale_factor, process_upscale],
            outputs=[output_res]
        )