HTML Accidental Mind [v3.5] [Myscra]

[DeanNorris]

The dev set up a patron page to commission some artwork, but couldn't find a reliable artist. As for generating images for the characters, that sounds good in theory, but generating decent images of pretty much any of the main cast besides Blaire can be difficult due to them having characteristics that are difficult concepts for image generators to understand. With Kiara many models barely distinguish between chubby and morbidly obese. Liane needs models that have decently diverse training sets or she will end up looking 60+ half the time. And holy crap, don't get me started on Prism (or Photona to a lesser extent). I felt like getting light effects originating from characters was nearly impossible even using some pretty unconventional methods I was messing around with a few months ago. I was considering tinkering with some of the newer tech to see if it was any better while playing the last update, but never got around to it.

I agree with you there.
Artwork will always look best when hand drawn by a talented artist, including every miniscule detail.
However, there are drawbacks. You mentioned it yourself, reliability is a factor, also a certain style might not be popular with some players.
Creating your own images gives the player the freedom to choose the style and fine-tune the looks of the characters until they fit their mental image.

But I also agree with you that current GenAI technology is still not quite there to do that easily and reliably, especially for more abstract looks. Getting the bodily proportions right for characters that have not typical model measurements (either fit, hourglass or voluptuous) can be tricky and requires quite some trial & error, but it is possible nonetheless. The same goes for the age. Adding some visual features in the definition to better fit the age the character is supposed to look like can help, I got some decent results for Liane that make her look like an attractive 45-year-old without sacrificing her looks or making her look 60+.

But yes, creating a character that looks close to what Prism is supposed to look like was a pain in the ass. I guess details that are very fantastic and otherworldly, so a lot of typical Superhero features, are hard to generate with such methods, simply because the pool of images the models have been trained on of course do not include models with superpowers like glowing eyes or RGB light-changing hair. So there are some hard limits that are difficult to recreate without some manual touch-ups in PS.

What's important to note here is that I am basing this on trying to create realistic looking images. In my head, the story is set in a somewhat realistic world, a point of reference in my head is the show "The Boys" which is played by real actors of course.
If someone would not be opposed to using anime characters, I imagine it would be much easier to include more fantastic features for characters that way. In general, the image pool for Anime seems to be more broad and detailed, however, I haven't fiddled with those art styles much so I wouldn't know for sure, just a heads up for anyone trying this out themselves.

In the end it comes down to personal preferences, I understand if someone feels the results are not high quality enough. I am personally happy and it adds to the enjoyment of the game.
I could theoretically share the results I have so far if someone was interested, but I'd rather not share AI generated "artwork" of the game without the Dev's consent.

 

[Fulminato]

How do you get Prism to level 3?
I don't see what to do next :

Spoiler

You don't have permission to view the spoiler content. Log in or register now.

Spoiler: Not a big spoiler, but you know...

You don't have permission to view the spoiler content. Log in or register now.

How do I do Nayika’s Level 2 Dreamwalk? I found the four islands and the sunken island and the island you can’t touch and the violins but that’s pretty much it and I don’t know where the key is.

you need to find the prison island (8,4) and remove the alteration here, the key is in the office

 

[HPLoveshaft]

Armory was throwing me off, I was thinking "Six alcoves are set into the eastern wall" was the clue.

enhanced vision is needed when you enter in the inner sanctum.
if you see the symbol in the room with the code it's an hint for the room you will find that number.

Spoiler: ligth Spoiler

You don't have permission to view the spoiler content. Log in or register now.

Spoiler: heavy spoiler

You don't have permission to view the spoiler content. Log in or register now.

Spoiler: Solution

You don't have permission to view the spoiler content. Log in or register now.

 

[zaxex]

What exactly do you need to do to get past the "gather the harem" step in the hints? I have gotten the call from Kiara and finished every thing else as far as I can tell.

Here's my save file if that helps.

 

[klckr]

Is it me or is this update a lot more buggy than previous ones? I tried 3.4a but it didn't do anything for me.
Liane is stuck at "gather the Harem". Do you need to have 3 or 4 girls at level 5 in the base?
Prism's investigation got stuck because when I go to La Luna (to get tickets) it loops back to "Meeting Nightingale" (at the Gala).
Nayika/Amita I never encountered, I don't know how to start her questline, is not in the hints.
I cannot build a lounge in my base as some people suggested.
Main quest got stuck before the costume, I had to nudge it in the variables.
The dev hints are not helpful at this point at all .

Edit: The nightingale questline I finished by going to Vivienne and asking about the Sefs and then going to the Tetons. This is from an investigation that I already did before but somehow it was available again and this closed the new one(?)

 

[KingAgamemnon]

Is it me or is this update a lot more buggy than previous ones? I tried 3.4a but it didn't do anything for me.
Liane is stuck at "gather the Harem". Do you need to have 3 or 4 girls at level 5 in the base?
Prism's investigation got stuck because when I go to La Luna (to get tickets) it loops back to "Meeting Nightingale" (at the Gala).
Nayika/Amita I never encountered, I don't know how to start her questline, is not in the hints.
I cannot build a lounge in my base as some people suggested.
Main quest got stuck before the costume, I had to nudge it in the variables.
The dev hints are not helpful at this point at all .

Nayika is started by getting peeped on at the library and at the mall, so do sexy stuff with your girls there and you'll eventually notice them.

 

[Mister Perfect Shadow]

I never played a text based only but is there no images of anything, how does this work?

 

[KingAgamemnon]

I never played a text based only but is there no images of anything, how does this work?

You read. Text-based only will describe the scene and the characters.

 

[littleowen]

I never played a text based only but is there no images of anything, how does this work?

Its not even a joke, you get a description of the events/girls and your head does the rest. Stick with it and you will start giving a unique voice to every character, i know i do

 

[GrumpyOlCoot]

So, an odd thing, on v.3.4, I can't seem to upgrade any of the bases. Dora sends the text mentioning it, but there's no option where it was. Not sure if it's a bug, or something changed between editions about it's location? Can attach the save if Dev wants it.

Edit: I feel like an idiot, the reason is because I didn't click Mark as Read on it, ignore me lmao

 

[KuroSama]

The dev set up a patron page to commission some artwork, but couldn't find a reliable artist. As for generating images for the characters, that sounds good in theory, but generating decent images of pretty much any of the main cast besides Blaire can be difficult due to them having characteristics that are difficult concepts for image generators to understand. With Kiara many models barely distinguish between chubby and morbidly obese. Liane needs models that have decently diverse training sets or she will end up looking 60+ half the time. And holy crap, don't get me started on Prism (or Photona to a lesser extent). I felt like getting light effects originating from characters was nearly impossible even using some pretty unconventional methods I was messing around with a few months ago. I was considering tinkering with some of the newer tech to see if it was any better while playing the last update, but never got around to it.

out of curiosity i decided to try generating an image out of kiara's description, this is my first result in a free AI site which is... not nearly as bad as i was bracing myself for.

 

[Penultimateanon]

out of curiosity i decided to try generating an image out of kiara's description, this is my first result in a free AI site which is... not nearly as bad as i was bracing myself for.

Ew. No.

 

[smkey21]

out of curiosity i decided to try generating an image out of kiara's description, this is my first result in a free AI site which is... not nearly as bad as i was bracing myself for.

That's the way these things go. To demonstrate what I'm talking about I tinkered with a short text prompt including the word chubby in it until I got something that was close enough to Kiara wearing her hoodie after she stops wearing a bra under it. I then generated 80 images from that prompt and looked through them for some of the more extreme examples. I used ReVanimated for this, mostly because it understands booru tags, which made getting the braless hoodie look pretty consistent. I've pasted together one of the slimmest, a more average, and one of the biggest outputs as examples of how widely the impact of the word chubby can vary within images generated from exact same prompt.

 

[AccidentalGryffindor]

That's the way these things go. To demonstrate what I'm talking about I tinkered with a short text prompt including the word chubby in it until I got something that was close enough to Kiara wearing her hoodie after she stops wearing a bra under it. I then generated 80 images from that prompt and looked through them for some of the more extreme examples. I used ReVanimated for this, mostly because it understands booru tags, which made getting the braless hoodie look pretty consistent. I've pasted together one of the slimmest, a more average, and one of the biggest outputs as examples of how widely the impact of the word chubby can vary within images generated from exact same prompt.
View attachment 3360576

She has a face like a toddler.

 

[KingAgamemnon]

She has a face like a toddler.

Anime artsyles lend itself quite heavily to neoteny.

 

[Jim Mann]

Is it me or is this update a lot more buggy than previous ones? I tried 3.4a but it didn't do anything for me.
Liane is stuck at "gather the Harem". Do you need to have 3 or 4 girls at level 5 in the base?
Prism's investigation got stuck because when I go to La Luna (to get tickets) it loops back to "Meeting Nightingale" (at the Gala).
Nayika/Amita I never encountered, I don't know how to start her questline, is not in the hints.
I cannot build a lounge in my base as some people suggested.
Main quest got stuck before the costume, I had to nudge it in the variables.
The dev hints are not helpful at this point at all .

Edit: The nightingale questline I finished by going to Vivienne and asking about the Sefs and then going to the Tetons. This is from an investigation that I already did before but somehow it was available again and this closed the new one(?)

You have to Kiara, Amita, and Crystal in the base. Then you should be able to call them all together during the Early Morning timeslot in the back half of the week. It's worth noting that 3.4 is bugged on this step. You need 3.4a someone provided on the last page.

 

[smkey21]

She has a face like a toddler.

Using cute in the prompt was probably the primary culprit. This was just to demonstrate an earlier point I made about the inconsistency most stable diffusion models have when generating characters using terms like chubby.

 

[AccidentalGryffindor]

Using cute in the prompt was probably the primary culprit. This was just to demonstrate an earlier point I made about the inconsistency most stable diffusion models have when generating characters using terms like chubby.

ReVanimated just doesn't understand why babies can't be sexy.

 

[DeanNorris]

That's the way these things go. To demonstrate what I'm talking about I tinkered with a short text prompt including the word chubby in it until I got something that was close enough to Kiara wearing her hoodie after she stops wearing a bra under it. I then generated 80 images from that prompt and looked through them for some of the more extreme examples. I used ReVanimated for this, mostly because it understands booru tags, which made getting the braless hoodie look pretty consistent. I've pasted together one of the slimmest, a more average, and one of the biggest outputs as examples of how widely the impact of the word chubby can vary within images generated from exact same prompt.
View attachment 3360576

I'm quite happy with the pictures generated by my prompts, though I preferred the RL photo style over anime.
As I mentioned earlier, it takes some tinkering to get it right, and obviously those pictures are not perfect or fully consistent, but it helped be immerse myself in the scenarios.

 

[AccidentalGryffindor]

I'm quite happy with the pictures generated by my prompts, though I preferred the RL photo style over anime.
As I mentioned earlier, it takes some tinkering to get it right, and obviously those pictures are not perfect or fully consistent, but it helped be immerse myself in the scenarios.

View attachment 3364119 View attachment 3364120 View attachment 3364121 View attachment 3364122

Love how the fourth one's just a pair of tits. Really captures her character.