Does "<" work with this version?

[Mistborn-First-Era]

I just started digging a bit and realized there was a new fork.
In the old original version "<" wouldn't be accepted and therefore I couldn't make a Lora wildcard inspiration. Has this been fixed in the latest version? I wanted to check before I updated and moved everything around.

[AlUlkesh]

Accepted where exactly? reading exif? Search? Custom paths?

[Mistborn-First-Era]

I made a .txt file with line seperated values such as

hairstyles.txt

bangs
top knot
pony tail
<lora:PinappleOnHeadHair:1>

It would error out when I used hairstyles.txt for the inspiration images script in A1111.

Hopefully that explains it better.

[AlUlkesh]

I just tried it out. It breaks here:

  File "C:\tools\Ai\stable-diffusion-webui\extensions\stable-diffusion-webui-inspiration\scripts\inspiration.py", line 59, in run
    os.mkdir(artist_path)

That's inside the inspiration extension, not the images browser.

The fix would be very minor, but yfszzx hasn't been seen since December.

I wonder if I should make a fork of that one too. You're obviously using that one.

[Mistborn-First-Era]

Ah sorry, I thought this was the inspiration extension fork. I guess Image browser is a fork of inspiration?
If it's a simple fix I would really appreciate it.
I looked into it with my limited python ability.
It looks like the problem is that it makes sub folders and then uses those folder names as prompts.

I did some debugging and the there are actually two problems that can probably be addressed at the same time.
1st: Since file names cannot have certain symbols

< (less than)
> (greater than)
: (colon - sometimes works, but is actually NTFS Alternate Data Streams)
" (double quote)
/ (forward slash)
\ (backslash)
| (vertical bar or pipe)
? (question mark)
* (asterisk)

Those cannot be used in your inspiration txt file. which is pretty limiting. I don't know if their would be a conflict in the ui as well within the "Inspiration tab -> get inspiration -> right side under get inspiration" since that uses the file name to allow for easy copy pasting.

2nd: This is probably a problem that originates from the 1st problem the more I think about it. But you cannot have "," (a comma) in your 'inspiration' prompt. For example

dog
cat
mouse
horse, farm, springtime
tiger, <lora:tigerlora:1>

would make the folders dog, cat, mouse, horse, and tiger; within each of which it would use only those words to generate the prompt. It cuts off farm, springtime, and <lora:tigerlora:1> from their respective lines.

Seems like it may be annoying to fix. I am pretty new to python still so I don't think I could fix it myself, but I think I did a good job debugging it.

[AlUlkesh]

Yes, I see. Well you could solve them both the same way by adding comma to that list of characters. And when you access paths you'd replace them by something else "_" for example. But keep the original data, so it can be used for the prompt and other things.

I'll look into it.

[Mistborn-First-Era]

I did it yay.

This is the code I have in the run function within the main class inside root->extensions->inspiration->scripts

It takes uses the first word of the inspiration.txt file as the folder name (it removes any illegal symbols and changes them to _)
then it takes the full prompt and uses that to generate the inspiration images.

for example:

dog
cat, woman, balcony
<lora:lion:1>, jungle, masterpiece
?????????pie???????????, curious

would make the folders

dog
cat
_lora_lion_1_
_________pie_________

Should work as the new run command.

       def run(self, p, prompt_placeholder, files):
        if not re.search(prompt_placeholder, p.prompt):
            raise Exception("Missing " + prompt_placeholder + " in prompt")
        original_prompt = p.prompt
        p.do_not_save_samples = True
        p.do_not_save_grid = True
        for file in files:
            tp = file.orig_name.split(".")[0]
            print("Process " + tp + "..")
            path = os.path.join(inspiration_dir, tp)
            if not os.path.exists(path):
                os.makedirs(path)
            f = open(file.name, "r", encoding='utf-8')
            line = f.readline()
            while len(line) > 0:
                line = line.rstrip("\n")
                # Ignore headline of artist.csv from automatic1111 repository
                if line == "artist,score,category":
                    line = f.readline()
                    continue
                double_quotes_match = re.search('^".+"', line)
                if double_quotes_match:
                    fixed_name = double_quotes_match.group()[1:len(double_quotes_match.group()) - 1]
                else:
                    name = line.split(",")[0]
                    fixed_name = ""
                    for letter in name:
                        bad_symbol_list = ["<", ">", "?", "\\", "*", ":", '"', "'", "/", "|", "?"]
                        if letter in bad_symbol_list:
                            fixed_name += "_"
                        else:
                            fixed_name += letter
                artist_path = os.path.join(path, fixed_name)
                if not os.path.exists(artist_path):
                    os.mkdir(artist_path)
                if len(os.listdir(artist_path)) >= opts.inspiration_max_samples:
                    print("Limit of " + str(opts.inspiration_max_samples) + " of '" + fixed_name + "' in " + tp + " is reached. Limit can be changed in the settings.")
                    continue
                p.prompt = re.sub(prompt_placeholder, line, original_prompt)
                print("folder name is", fixed_name, "using", p.prompt, "as the prompt.")
                line = f.readline()
                processed = processing.process_images(p)
                for img in processed.images:
                    i = 0
                    filename = os.path.join(artist_path, format(0, "03d") + ".jpg")
                    while os.path.exists(filename):
                        i += 1
                        filename = os.path.join(artist_path, format(i, "03d") + ".jpg")
                    img.save(filename, quality=80)
        return processed

ps. how do I fork myself? I would like to get a green dot on my github record I can use to help land a job once I am done with school.

[AlUlkesh]

Nice 😃

You can fork here, top of the page:

Small tip: regex is really good for replacing those values, for example:

invalid_chars = '<>:\"/\\\\|?*\0'
fixed_name = re.sub(f'[{re.escape(invalid_chars)}]', '_', name)

[Mistborn-First-Era]

Think I did it. Thanks for the help. I will have to learn regex. I tried, but it was hard to understand.
https://github.com/Mistborn-First-Era/stable-diffusion-webui-inspiration-Functioning-Loras