How to do batch img2img job of many img/prompt pairs (Solved with one question)

[yiouyou]

Hi,

I need to re-draw 100 images. Thanks to clip-interrogator, I've generated prompt text for each one of them. Next, I should to run img2img. However, I noticed that you cannot sigh the prompt for each image specifically with img2img batch tab.

I tried to write a script about it as below (After a few days' debugging, it's working now):

import modules.scripts as scripts
import gradio as gr
import os
import random
from PIL import Image, ImageOps

from modules import images, processing, shared
from modules.processing import process_images, Processed
from modules.shared import opts, cmd_opts, state


def load_prompt_file(file):
    if file is None:
        lines = []
    else:
        lines = [x.strip() for x in file.decode('utf8', errors='ignore').split("\n")]
    return None, "\n".join(lines), gr.update(lines=2)


class Script(scripts.Script):

    def title(self):
        return "1_img/1_prompt img2img"

    def show(self, is_img2img):
        return is_img2img

    def ui(self, is_img2img):
        different_seeds = gr.Checkbox(label='Use different seed for each picture', value=True, elem_id=self.elem_id("different_seeds"))

        prompt_txt = gr.Textbox(label="List of prompt inputs", lines=1, elem_id=self.elem_id("prompt_txt"))
        prompt_file = gr.File(label="Upload prompt inputs", type='binary', elem_id=self.elem_id("prompt_file"))
        prompt_file.change(fn=load_prompt_file, inputs=[prompt_file], outputs=[prompt_file, prompt_txt, prompt_txt])
        prompt_txt.change(lambda tb: gr.update(lines=2) if ("\n" in tb) else gr.update(lines=2), inputs=[prompt_txt], outputs=[prompt_txt])

        img_input_dir = gr.Textbox(label="Input directory", elem_id=self.elem_id("img_input_dir"))
        style_token = gr.Textbox(label="Fine-tuned model style token", elem_id=self.elem_id("style_token"))
        return [different_seeds, prompt_txt, img_input_dir, style_token]

    def run(self, p, different_seeds, prompt_txt, img_input_dir, style_token):
        prompt_lines = [x.strip() for x in prompt_txt.splitlines()]
        prompt_lines = [x for x in prompt_lines if len(x) > 0]
        img_output_dir = img_input_dir + "-out"
        if style_token == '':
            prompt_end = ''
        else:
            prompt_end = ' , '+ style_token
        p.do_not_save_grid = True
        p.do_not_save_samples = True
        p.n_iter = 1

        image_files = shared.listfiles(img_input_dir)
        image_files_num = len(image_files)
        print(f"Will process {image_files_num} images, creating {p.batch_size} new images for each.")

        # print(len(prompt_lines), image_files_num)
        assert len(prompt_lines)==image_files_num

        state.job_count = image_files_num
        for i, i_image in enumerate(image_files):
            print(f"{i+1}/{image_files_num} image: {i_image}")
            if state.skipped:
                state.skipped = False
            if state.interrupted:
                break
            i_img = Image.open(i_image)
            i_width, i_height = i_img.size
            p.width = i_width
            p.height = i_height
            i_img = ImageOps.exif_transpose(i_img)
            i_prompt = prompt_lines[i] + prompt_end
            i_neg_prompt = "blurry, ugly, out of focus, (((out of frame)))"
            p.init_images = [i_img] * p.batch_size
            seeds = []
            for b in range(p.batch_size):
                seeds.append(int(random.randrange(4294967294)))
            p.seed = seeds
            p.prompt = [i_prompt] * p.batch_size
            p.negative_prompt = [i_neg_prompt] * p.batch_size
            p.prompt_for_display = i_prompt
            processed = process_images(p)

            for k, processed_image in enumerate(processed.images):
                left, right = os.path.splitext(os.path.basename(i_image))
                _png = f"{left}-{i}{k}.png"
                _txt = f"{left}-{i}{k}.txt"
                os.makedirs(img_output_dir, exist_ok=True)
                _png_path = os.path.join(img_output_dir, _png)
                _txt_path = os.path.join(img_output_dir, _txt)
                processed_image.save(_png_path)
                with open(_txt_path, "w", encoding="utf8") as _txt:
                    _txt.write(processed.infotext(p, k))

        return None

One left issue:

1. All though the script read the image from the input folder, it still requires to upload a init image for the img2img tab. Otherwise, there will be some error reported. Any one know how to point the init image from the input folder automatically by the UI?

Thanks for your help and attention~

[Novatron-AI]

parser.add_argument(
"--init-img",
type=str,
nargs="?",
help="path to the input image"

This is how img2img.py gets an init image