Futari Bus - Vol. 1 Ch. 11.5 - Volume 1 Extras

Dex-chan lover
Joined
Jan 20, 2025
Messages
6,046
it seems like there will be a sequel where MC and An-chan going to isekai :aquadrink:

10XNRom.png


thanks for bonus
 
Dex-chan lover
Joined
Dec 3, 2023
Messages
2,147
What do you mean "swords not included in the story?!" I thought this was manga??? Like Goku and Death Note.
 
Dex-chan lover
Joined
Sep 30, 2023
Messages
575
first i've seen those anti ai watermarks, i wonder how well they work
A ton of artists who post to Twitter are using them these days, since Twitter uses everything posted to train grok now. Unfortunately, my hunch is that they probably don't work very well.

Given that AI can be trained to recognize and filter other features in art, there's no reason to think it can't do the same for watermarks. It's not the early days where signatures and watermarks could leak into generated images simply because it didn't understand what they were. I imagine current models will just end up aware that "anti AI watermark" is in the image, and otherwise understand and catalogue all of the other features of said image without issue as though the watermark wasn't even there.

Heck, I'm pretty sure models already exist that can simply remove the watermarks entirely and present to you the pristine version.

AI poisoning is where the real good shit is. Unfortunately it's not super accessible yet, to my knowledge at least, and I'm not sure about its cross-model effectiveness.
 
Last edited:
Dex-chan lover
Joined
Jun 28, 2025
Messages
3,271
A ton of artists who post to Twitter are using them these days, since Twitter uses everything posted to train grok now. Unfortunately, my hunch is that they probably don't work very well.

Given that AI can be trained to recognize and filter other features in art, there's no reason to think it can't do the same for watermarks. It's not the early days where signatures and watermarks could leak into generated images simply because it didn't understand what they were. I imagine current models will just end up aware that "anti AI watermark" is in the image, and otherwise understand and catalogue all of the other features of said image without issue as though the watermark wasn't even there.

Heck, I'm pretty sure models already exist that can simply remove the watermarks entirely and present to you the pristine version.

AI poisoning is where the real good shit is. Unfortunately it's not super accessible yet, to my knowledge at least, and I'm not sure about its cross-model effectiveness.
I know Glaze has been a thing for awhile, and isn't the most confusing to use, though time-intensive as all worthwhile things can be.
 
Dex-chan lover
Joined
Sep 30, 2023
Messages
575
I know Glaze has been a thing for awhile, and isn't the most confusing to use, though time-intensive as all worthwhile things can be.
Yeah, it's just unfortunate that the tech we have available to do this right now is just as power hungry and resource intensive as the tech it's trying to defeat. And it's kinda difficult to gauge its actual effectiveness. But gotta do what we can I guess.
 
Dex-chan lover
Joined
Feb 19, 2018
Messages
107
first i've seen those anti ai watermarks, i wonder how well they work
Nowadays, it doesn't work at all. Back when the edit image feature is introduced in Xwitter, an artist made a post to show how clever the AI is. It can remove watermarks, complete a sketch, redraw, color etc.
So basically, as of now, the watermark is put there as a sign of consent. If the AIbros don't care, oh well.

The closer school for some people made them lazy or careless, and lead them being late to school.

A new journey, sometimes it starts with a farewell. And our Futaba's journey started with the vending machine going away. :salute:

And the journey of Shunpei realizing he has feeling for Ann started with his shoe going away....
 

Users who are viewing this thread

Top