You need to have them in square brakcets to ‘run’ them:

{[subAgilityMod = 1, '']|[subAction = advnonac, '']}
// The (, '') just tells the square bracket to output nothing.

Also if you are calling a list inside a square bracket, you don’t need to enclose it too in another square bracket.
[subAction = advnonac]✅.
[subAction = [advnonac]]❎.

You can use the select-until-plugin instead of filter-list-plugin.

selectUntil = {import:select-until-plugin}

count = 0

output
  [count = 0, acronymLoop]
  // count = 0, to reset the counter for each generation
  
acronymLoop
  [word = count < acronym.length ? (n = selectUntil(noun, i => matchFirstChar(i, acronym.charAt(count))), count++, "<span contenteditable>" + emphasizeFirstChar(n) + "</span> " + acronymLoop) : '']

  // Instead of filter list, we use the select-until-plugin in which the 'test' to be used was the `matchFirstChar` function that you already have.
  // We start at 0, then count up (`count++`)
  // Then the check is now `count < acronym.length` instead of `count <= acronym.length` since we use zero-indexing with `charAt`.

You have the items in gender to be male, female, nonbinary, etc., and not Male, Female, Nonbinary in which you are checking male to Male which is not equal/same, you need to have the dynamic odds check lowercased.

Lesbian^[g == "female" || g == "nonbinary"]
Lesbian^[g != "male"]

With this, you can also specify directly each items’ odds:

output
  Race: [r = races.selectOne, r.getName] <br> Sub-Race: [r]
  Race: [r = races.evaluateItem] <br> Sub-Race: [races[r]] // Using the Dynamic Sublisting.

races
  Human^1.5
    Eastern Human^0.4
    ...
  Elf^0.2
    High Elf^0.3
    Wood Elf^0.3
    Dark Elf^0.1
  ...

This would mean, you can have the chance of it being a ‘Elf’ or other races, then you can also specify the odds of the subraces.

Please see this generator to see each effect of the ‘text’ that on the prompt.

I’ve separated the text into sections, to see the effect of adding that text to the generated image. All images use the same seed, so that it would have the (almost) same noise to generate from. With or without your provided photo (on my device), the result was still the same.

Some major effect on the image generated is the addition of the facial hair text on the prompt.

The more interesting thing would also be I could adjust the weight of it to 1:5 and the effect would be more significant

Yes, if you increase the ‘emphasis’ of a tag, the AI would increase the significance of that text in the image. See Tag Emphasis - Prompting Guide for a demo.

Based on your response to me before, it seems the image actually changes and poisons the image pool.

If you use popular people’s names (Trump, Elvis, etc.) the images generated wouldn’t be as accurate/realistic. The reason might be to prevent ‘deep fakes’ which may be used to impersonate/falsify information (this is not confirmed since we don’t know the specific model used in the images and we don’t know the images/dataset used to train it).

Can you also provide the prompt of the generated image? Click the ℹ on the top left to open the prompt and negative prompt of the generated image like so:

It might just be a coincidence that the prompt specifically gives that image based on that file path. And as I said, if the page was able to open it up, it would have added cat features to it regardless if it has the ‘a man referred to’ as a starting prompt since it should’ve been able to infer the image was a cat and it should add cat features since it is a reference to it. I would suggest not to post any personal info for privacy.

Also it doesn’t train on the spot. It is already trained and it would just output what it has been trained on. And as you said, you added the a man referred to ... to determine the following traits of char1: facial structure, head shape, proportions, colors, detail, moles, hair style, facial hair, eye shape, health, hair line to emphasize that it was a man, and with these texts it would more likely that the images would lean into it.

It is just added as a text in the prompt: Here, I’m using an image of a kitten as the input on the filepath. The filepath was /storage/emulated/0/DCIM/Camera/IMG_20231219_100845102.jpg

And as you can see, by clicking on the (i) button on the top left, you can open the prompt that is used to generate the image.
Of course the image would be different since there are different text on the prompt, also images are different due to the seed. But, it should’ve had a cat features, or cat related images since the file path was pointing to a cat image. So it cannot open the image and infer to it. It is only a text-to-image not image-to-image.

Here is with the same filepath that you have: /internal storage/dcim/perchance/Me.jpg which I don’t have on my device:

Looking at your prompt:

, ,a man referred to as "char1". (char1 is the person in the local image "/internal storage/dcim/perchance/Me.jpg": 1.3). reference "/internal storage/dcim/perchance/Me.jpg" to determine the following traits of char1: facial structure, head shape, proportions, colors, detail, moles, hair style, facial hair, eye shape, health, hair line

Since you have the words facial structure, head shape, proportions, colors, detail, moles, hair style, facial hair, eye shape, health, hair line the AI would try to generate an image with those words.

With the text (char1 is the person in the local image "/internal storage/dcim/perchance/Me.jpg": 1.3) it would have an increased emphasis by 1.3 on the generated image. So, if you have a filepath with samus on it, the AI would greatly emphasize those words, meaning it would appear emphasized or will appear most likely on the image.

The file path is just treated as a text input.

It is because the AI model is trained on images of Samus from Metroid from the get go. Regardless of the images of Samus on your device. It doesn’t look into your device, if it would it would be a breach of privacy. Any famous/familiar/popular characters would be able to be generated albeit with variations and differences.

Yes, the Perchance would only parse them as a string, meaning literal text of the file path. It wouldn’t open them, it would only infer on the ‘text’ provided. I would just remove those ‘paths’ on the prompt and see if that works.

Can you link the generator page?

If you are using a text-to-image-plugin, it would always start to generate immediately after page load. Also, you cannot use local files in the text-to-image, it would only translate as a file path, in which case, it would get the file path as a prompt. If your file path has a ‘hair style’ text on it, then it would be the reason why there are hairstyles on the image generated.

I’ve tried testing it with [console.log(com), update(), console.log(com), ''] upon first load of the page and update essentially removes the com variable since it cannot update the comments instance.

Based on the note on the comments-plugin page:

Note: By default comments areas aren’t “updated”/randomized when the user clicks the “randomize” button. Instead the comments area will “stay there”. If you actually want to display a random channel each time the user clicks a button, or you e.g. want to put a different comments box in each “room” within your goto-plugin adventure, then you need to add replacedDuringUpdate=true like in this example generator. That ensures that the old comments section is “deleted” and replaced by a fresh one every time the page is updated/randomized.

With the replacedDuringUpdate=true option, you can then update the comments with it and the com would still be accessible. Otherwise, you need to set the comments instance in the com again if you update the page.

settingsButtonStyle, fullscreenButtonStyle were added a while back, submitButtonSuccessText, hideSettingsButton, hideFullscreenButton were quite new I think.

Pinging dev @perchance@lemmy.world

You can possibly use the API that retrieves the generator’s stats for the generator cards. Though you need to make the function an async function.

async loadUserDataset() => // Need to change it to an Async function to wait for the data to be retrieved.
  ...
  if(userDataset != "") {
    let result = await fetch("/api/getGeneratorStats?names="+userDataset).then(r => r.json()); // Gets the Generator Stats
    let gens = result.data; // Returns an Array of Generator
    if (gens.length > 0) { // If there is data
      document.userDataset = userDataset;
    } else { // if not
      alert('Perchance Page Not Found');
      document.userDataset = 'template-fusion-generator-dataset' // or any default pages
    }
  }
  ...
  update() // to refresh the display, or you can use element.innerHTML = this new value

I’ve updated the code on the demo that I’ve posted. It now should only load the Iframe when the user clicks on the load comments.

The only problem I’ve encountered was that if you have another instance of the comments i.e. same channel already loaded in the page, the load comments wouldn’t work. If you have different channels though, it works.

Here is a generator that might be something useful: https://perchance.org/e3gnujedx3

Essentially it is a checkbox that saves the ‘state’ of the checkbox in a localStorage.

Then on the onLoad() of the comments option, I’ve added a check there to see if the saved state is enabled or not. Then I added a button to load the comments if it isn’t enabled on load.

Since you are using a document.getElementById(...).click() on the button with the openTab HOWEVER, the openTab function is not yet defined since it is under the click() which is why it throws that error. So you just need to use the click() below the declaration of the function. I would also not use this as an Id of an element since this has a different use in JavaScript.

<script>
  function openTab(evt, tabName){
    var i, tabcontent, tablinks;
    
    tabcontent = document.getElementsByClassName("content");
    for (i = 0; i < tabcontent.length; i++) {
      tabcontent[i].style.display = "none";
    }
    
    tablinks = document.getElementsByClassName("topbtn");
    for (i = 0; i < tablinks.length; i++) {
      tablinks[i].className = tablinks[i].className.replace(" active","");
    }
    document.getElementById(tabName).style.display = "block";
    evt.currentTarget.className += " active";
    
  }
  
  document.getElementById("TabOneBtn").click(); // click after declaring the function, changed the id of the button from 'this' to 'TabOneBtn'
</script>

Another possible way to fix your default is to probably just add the active class on the default one, and no need to click the button to set the default.

Some update to this:

• I’ve updated converter URL: to /seventh-sanctum-converter.
• The // SEED based conversion seems to be working, though there are manual checks to make sure the generators are working properly.
• The Data Archives might not be updated, though some of the gens in the Data Archives are no longer available in the site as well.
• I’ve compiled my conversions at /7th-sanctum-conversions-vionet20.

There are some characters made by users at the Community Made Guide for the AI Character Chat. I would recommend checking out the GPT Roleplay Realm characters.

If you want to run some code but ignore any Perchance errors that occur during it, then use this: window.ignorePerchanceErrors(() => { putYourCodeHere }). The return value of this function is the return value of the callback code that you provide. Useful when evaluating user-provided Perchance syntax which may have bugs. You can also run window.clearPerchanceErrors() to clear all existing error logs.

From the Advance Perchance Tutorial