discussion Anti-Social & Lurker - Hideout spot

[Redyscom]

I think you're taking my dumb swipe way too seriously

There's nothing wrong with taking it seriously—that's how good memes are born... I don't know why I wrote this.

 

[Yoriko]

I think you're taking my dumb swipe way too seriously

English is not my native language, it’s hard to pick up on sarcasm, irony, or jokes hahaha, my bad.

 

[CubbyBuddy]

English is not my native language, it’s hard to pick up on sarcasm, irony, or jokes hahaha, my bad.

Nah, it's fine. It's hard to pick up sarcasm in text in general

 

[Lustre]

You must be registered to see links

Just found out Perchance has multiple image generators!

 

[Redyscom]

English is not my native language, it’s hard to pick up on sarcasm, irony, or jokes hahaha, my bad.

Yeah, I totally get that.


Never underestimate the power of the horny little bunny!
(She just wiped out a boss's entire team without even getting hit.)

You must be registered to see attachments

 

[Yoriko]

Just found out Perchance has multiple image generators!

Yeah, it helps a lot when you don’t have a local generator, because it’s pretty unrestricted, but I’ve never managed to generate something 100% good on it, have you?

 

[dimodamo]

I see it differently. I started by generating images on

You must be registered to see links

, refining them in Nana Banana, and then making some edits by hand. They always turned out terrible, and even so I still won a few times.


I don’t really know what Mime’s criteria are, but I always try to tell a story with the images I generate and actually put time and effort into them. Even when they don’t turn out great, they still feel worth it because it’s fun. For me, that’s the main point. The prize is just a bonus.

Same, I can't just leave it on just one image, my brain just tells me to create a secuence of images, after all half the fun of using perchance is iterating the prompts until you get something good, even with the consistency problems perchance has, and who knows, sometimes I get an image that while it wasn't what I was looking, it's so good I add it to the secuence anyway.

 

[Lustre]

Yeah, it helps a lot when you don’t have a local generator, because it’s pretty unrestricted, but I’ve never managed to generate something 100% good on it, have you?

I don't know what you'd consider completely good, but I've got some pretty decent stuff out of it. I've found the key is to keep it simple (1-2 people, maybe a background) and have a pretty clear vision but not on the result (idk if that makes sense). It might take a while, but I'll eventually get what I want. (I attached two sfw examples I made with perchance)

 

[Yoriko]

Yeah, I totally get that.


Never underestimate the power of the horny little bunny!
(She just wiped out a boss's entire team without even getting hit.)

You must be registered to see attachments

You must be registered to see links

 

[dimodamo]

I don't know what you'd consider completely good, but I've got some pretty decent stuff out of it. I've found the key is to keep it simple (1-2 people, maybe a background) and have a pretty clear vision but not on the result (idk if that makes sense). It might take a while, but I'll eventually get what I want. (I attached two sfw examples I made with perchance)

Yeah, 3 or more non-generic people and perchance just craps itself, either adding more people that you asked and/or mixing their traits.

 

[Yoriko]

I don't know what you'd consider completely good, but I've got some pretty decent stuff out of it. I've found the key is to keep it simple (1-2 people, maybe a background) and have a pretty clear vision but not on the result (idk if that makes sense). It might take a while, but I'll eventually get what I want. (I attached two sfw examples I made with perchance)

No way! How did you manage that? Every time I try to make something there, I sometimes have to fix the eyes or correct some anatomy.

It turned out really good! How do you make your prompts?

 

[Lustre]

Hands and stuff that are not common (or even extant) in the training images will also mess you up. That's why for more "extreme" content the anime models tend to be better (I assume because of hentai?) and can be a good testing ground for the prompt if you want to turn it realistic, however, if there are no example of that in the training images, it gets confused. I'd say the 40s, 50s, 60s, etc image generators are good examples of this.

Yeah, 3 or more non-generic people and perchance just craps itself, either adding more people that you asked and/or mixing their traits.

Also perchance is effectively "old tech" compared to more advanced models, maybe a year or two behind. It's a you get what you pay for imo.
[automerge]1773949251[/automerge]

No way! How did you manage that? Every time I try to make something there, I sometimes have to fix the eyes or correct some anatomy.

It turned out really good! How do you make your prompts?

eyes are super annoying. I've had perfect pics I've had to throw out because of the eyes (or extra limbs).

 

[Redyscom]

With all this talk about AI generation, it's reminding me of Nvidia's announcement about DLSS 5, which is basically a slop filter.
If that's what AAA games will be like, no one's going to take me away from indie games, independent AA titles, and fan games.

 

[Lustre]

It turned out really good! How do you make your prompts?

this is what I used for some of what I posted in Mime's thread yd:
streamer gamer girl, she is wearing sexy clothes, sexy body, hot, wearing a headset, she is try to look her best for her stream and garner donations, cosplaying, cute face, natural face, legs spread, skirt riding up, hand covering her privates

then the negative prompts are always something like:
wrong limbs, too many limbs, fat, ugly, bad body (parts or whatever)

() and simple directions between the commas are super important since it treats each one sep and changes the order of operations and stuff.

 

[Yoriko]

With all this talk about AI generation, it's reminding me of Nvidia's announcement about DLSS 5, which is basically a slop filter.
If that's what AAA games will be like, no one's going to take me away from indie games, independent AA titles, and fan games.

I think it’s unanimous, no one liked it hahaha. I’d like to know, what goes through the mind of a multi-million dollar company to rely on such a shitty idea of turning any game into AI slop.

 

[Redyscom]

I think it’s unanimous, no one liked it hahaha. I’d like to know, what goes through the mind of a multi-million dollar company to rely on such a shitty idea of turning any game into AI slop.

It's terrifying to think that people with so much money are so stupid and so completely disconnected from reality.

As I was writing that, I remembered how the stock prices of several video game companies plummeted the day Google demonstrated an AI capable of creating hideously ugly walking simulators based on images and prompts.

They think we're still stuck in the Pac-Man and Tetris era, believing that those “slop-games” have a real future, while ignoring the fact that games require millions of lines of code just to function even mediocrely.

 

[redfingers]

I see it differently. I started by generating images on

You must be registered to see links

, refining them in Nana Banana, and then making some edits by hand. They always turned out terrible, and even so I still won a few times.


I don’t really know what Mime’s criteria are, but I always try to tell a story with the images I generate and actually put time and effort into them. Even when they don’t turn out great, they still feel worth it because it’s fun. For me, that’s the main point. The prize is just a bonus.

Your creations have looked great to me! I will usually try to add something ridiculous to compensate for my lack of using powerful image generators (so far)

 

[Yoriko]

this is what I used for some of what I posted in Mime's thread yd:
streamer gamer girl, she is wearing sexy clothes, sexy body, hot, wearing a headset, she is try to look her best for her stream and garner donations, cosplaying, cute face, natural face, legs spread, skirt riding up, hand covering her privates

then the negative prompts are always something like:
wrong limbs, too many limbs, fat, ugly, bad body (parts or whatever)

() and simple directions between the commas are super important since it treats each one sep and changes the order of operations and stuff.

That’s cool, you mix some “tags” with more descriptive parts (I’m not exactly sure how to describe your prompt style, I hope you understand me, because I don’t even understand myself hahaha).

Maybe keeping it simpler is better, my prompts were all more descriptive, like I was telling a story.
I delete my prompts after making the images D: so I can’t really show them.

Your image generation turned out really good, I’ll take note of it here, thanks <3

 

[CubbyBuddy]

It's terrifying to think that people with so much money are so stupid and so completely disconnected from reality.

Remember that Nvidia has invested very heavily in AI, so part of this is sunk cost

 

[Redyscom]

Remember that Nvidia has invested very heavily in AI, so part of this is sunk cost

And I'm really glad they're suffering the consequences of their own stupidity. What I don't like is that I know the ones who will pay the price are innocent workers, not the high-ranking officials who made such bad decisions—as always.