"Creativity Transfer" implementation

[sharp-trickster]

https://github.com/zyxelsa/creativity-transfer

Just another cool thing to hopefully make the integration list on this interface.

[Ehplodor]

I was hyped but, what is the difference with currently available textual inversion ?

[some9000]

_{^{Why, by the way, does everyone simply ignore the indication of the minimum system requirements.}}

What are you talking about? Shouldn't SD just run on a 10 year old smart toaster?

[Ratinod]

What are you talking about? Shouldn't SD just run on a 10 year old smart toaster?

SD can. But DreamBooth and some other training "things" not.

[papersplease]

The difference is one-shot learning and better extraction of relevant stylistic features, it seems from the paper.

[sharp-trickster]

Reading the paper I was under the same impression as @papersplease.
I think there´s A LOT going on WAY TOO FAST in AI in general, lots of papers on a daily basis, and sometimes real gems fly under our radar for a while. My intention bringing some of this papers here to the "ideas" section is to bring them to a "local peer review", as in: People smarter than me weighting in if that could be an interesting thing to implement on this GUI anytime in the future.
I hope I´m not disrupting the work done here in any way by doing so, and I´m ok to not do so anymore if I am.
Thank you for checking it out.

[Ehplodor]

@sharp-trickster it's OK really. I upvoted because I think so too. Really happy to discover thank to you yet another paper about this. I just wanted to emphasize that the functionality proposed in this paper (one-shot learning style transfert as textual embedding) is implemented into SD webui as the DreamArtist extension for some weeks (months ? see closed PR link) already.