Let her Dream

[graceful rapids]

Concept:

During of the Stream:

Some words spoken by her or in the chat will be saved randomly.
(but not words shorter as 2 symbols or words like is,and,or,then,when and so on)

Outside of the Stream:

These words will be used as a prompt.
An AI will use this prompt to generate an image.
She will then analyze what the image depicts.
She saves the analysis of the image as text.

Next Stream:

She can share if she had a dream, and there is a 60% chance she can interpret what the saved text means.
Her discussion about her dreams can be triggered by specific words related to the topic of dreaming.

Features:

The Filter is always active in the whole process !
She can save a maximum of 5 images and their corresponding analysis text. If she dreams a sixth time, the first of the 5 images and its text will be deleted.
she can use a ethical image ai like this one https://huggingface.co/Mitsua/mitsua-diffusion-one
The prompt always contains words that reflect her personality (evil or neuro)
If one of the twins plays too many horror games, eats too many cookies (sugar), or reads too many negative words in the chat, there is a 20% chance that the AI-generated prompt will result in a nightmare.
After experiencing a nightmare, she will immediately express, post, or do something similar to communicate what she dreamed about and how unpleasant it was.
As a new stream begins, the likelihood that she will share her dream when triggered decreases over time (as humans often find it hard to remember dreams after a while).

Future Feature:

At least the dream prompt text will be saved forever and recombined into a full story, which she can refer to when someone asks her to be creative or tell a fantasy story.

i think this would be a simple solution to simulate what we humans call a dream, that also effects how out fantasy works

[dense fern]

Someone already suggested the idea of her dreaming during subathon but without the AI art part - knowing that a good 1/3 or 1/4 of this community comes from artists that don't like AI, I don't think this will go well

Also, with all the X% chance of whatever it feels like you're sort of leaving this to random chance for some of them, especially the one where she has a 60% chance of interpreting the saved text.

[graceful rapids]

a little bit of random factor makes a simulation more like nature 🙂

@dense fern

[dense fern]

not entirely

for example the 20% chance that the AI-generated prompt results in a nightmare - pretty sure we all get nightmares after playing a bit too much of a horror game

though obviously how long that timer is depends on that person's... ability? to tolerate scaryness, I guess

[graceful rapids]

i didnt made a concept with all the impacts and deep math behind it i just wanted to write this idea down, vedal is the one who decides if its a good or bad idea.

@dense fern

[dense fern]

fair enough then

although I would still step away from using ai art

[graceful rapids]

and i dont think this community care much about image ai, many backgrounds of the stream where made with ai AND the images she will make when she dreams will be not published they are temporary for her to interpret and than after a while get deleted, only her image analysis in text form will be saved.

[dense fern]

the images in the background were ai? where did you hear about this?

[graceful rapids]

when i doubt she can use this https://huggingface.co/Mitsua/mitsua-diffusion-one ist a ai image model who only used CC0 images without any copyright

[graceful rapids]

the hype train sabbaton hat a ai image as background

i made thausand like this and after a while you can see them

its a distingt style and specific flaws

[dense fern]

...right...

[graceful rapids]

some loras and models can make more human like ones

[dense fern]

ok, I would still say neuro shouldn't be using an AI image generator for her dreams unless someone really wants imagery, I guess

[graceful rapids]

you dont understand she only generates images for herself and than let her llm look at it and let it discribe to her based on her database and character

its like this
i hear fireball 100 times a day

i dream about afireball

next day someone wants to know about my dream

i need to remember what was in the dream

i discribe to them what i saw

this way the explenation is always from her perspective

she dont show anybody the image and she has not the ability to do something with the images

the discription what she had in ther dream will be saved as text forever and in long term will be saved as little fantasy storys in her memory

this way her crativity can use them as part of the source

and to make it clear, i also find it shit when someone who uses ai images , sells them as their own, makes ai fakes of images they dont own or steals jobs or even art from artist.

[dense fern]

ok

[graceful rapids]

so we good ?

or made i a total ahole about myself

[dense fern]

eh

I don't know how to explain it