A couple of questions

[juan9999]

Firstly, I have to say great work!

I've been struggling with long prompts that are difficult to tweak and I think this is the solution to those just by looking at your quick start guide!

Currently I use improved prompt matrix with very long prompts with multiple matrices so it definitely would a huge step towards simplification and more automation!

A few questions, as I begin to understand the possibilities:
1- does it do every permutation of options as improved prompt matrix or is it mostly random (which is fine too)

2- What I'm currently struggling with is workflow;

Currently I generate a lot of images at 512x512 using SD 1.5 and then go through them and create a folder of images that I like (that have the prompts in the png)

Can I somehow use your solution to batch process these png files to go back and use the prompts stored in the png to regenerate them at a larger size (say 1024x1025) with hi-res fix and a diffusion of [.3|.35|.4] to get the same images larger then feed them to img2img to use AI to enlarge them? If so how would I go about it?

Thanks so much!

[ThereforeGames]

Hi @juan9999,

Thanks for reaching out!

I released an update for Unprompted a few hours ago (v4.2.0) with a couple new features that should help you reach your goals.

In response to your questions directly:

1. At the moment, Unprompted does not have an easy way of producing all permutations. However, some shortcodes like [choose] can be given a _case value which allows you to override the randomness with a predetermined outcome. You can combine this with the batch_index variable which corresponds to your progress within a batch run.

2. For this, I recommend looking at the new [filelist] and [length] shortcodes.

Here are some thoughts on how I would structure the template:

[set my_files][filelist "C:/my_pictures/*.*"][/set]

This will store your list of files into the my_files variable, e.g. "C:/my_pictures/picture_one.png|C:/my_pictures/picture_two.png|C:/my_pictures/picture_three.png|..."

You can then retrieve the number of files, which we can set to our SD batch count:

[set batch_count][length "{get my_files}"][/set]

Now, we can select an individual file for processing with batch_index, which corresponds to our progress in the batch run:

[set this_file][choose _case="{get batch_index}"][get my_files][/choose][/set]
[init_image "{get this_file}"]

Finally, we need to extract our prompt from the full filepath. It's a bit cumbersome, but I think you can use a couple instances of [replace] for this:

[set prompt][replace _from=".png" _to=""]{replace _from="c:/my_pictures/" _to=""}{{get this_file}}{/replace}[/replace][/set]

EDIT: I think [filelist] might use a backslash instead of a forward slash somewhere, like "C:/my_pictures\example.png." You should probably turn on debug mode in config_user.json and monitor the console to see exactly how these paths are formatted.

That should give you "my prompt words" from "C:/my_pictures/my prompt words.png".

Anyhow, this is all theoretical and may take some trial and error to get working. I hope this was enough to help you get started though. Please let me know if you run into any obstacles.

Happy Holidays!