Street Fighter-1950s Super Panavision 70 AI trailer

Рет қаралды 143,851

Gem Studio

Күн бұрын

Пікірлер: 43

@NeverMadeFilms 20 күн бұрын

Dope !

@geraldbrown2893 2 ай бұрын

Zangief looking straight up rip 😲 though I love seeing how attractive Chun-Li looking right here 😍🥵❤️👌🏿

@spidey0123 2 ай бұрын

Like this version of Ken and Ryu better than other ai SF videos here.. I thought Chun Li was gonna dance. Lol

@unexpectedlyincognito 2 ай бұрын

1:03 blanka is cool looking Bad ass vid man 🤘

@Zapa-pd6sw 2 ай бұрын

Yea more like lunatic in the jungle 🎉🎉 very intimidating to me

@isaacorlich 2 ай бұрын

no Cammy is a crime...

@delanteawashington8910 10 күн бұрын

Ken is my top favorite

@ConcoctInc 2 ай бұрын

0:09 finger

@bastiancrack9802 2 ай бұрын

Primer comentario dé septiembre 2024 👍

@Buddy_Kryyst 2 ай бұрын

Fantastic

@Gemstudio-r1w 2 ай бұрын

Thanks

@jcaro5954 2 ай бұрын

This looks better then Snow White

@Gemstudio-r1w 2 ай бұрын

Of course.but I will make Snow White soon

@АнатолийГоловач-ы3р Ай бұрын

❤👏💪👍🤝

@robinly 2 ай бұрын

M. Bison looks like Danny Huston!

@JyGproductions 2 ай бұрын

Some peoplein the back ground have their heads reverse.

@alexstone7498 2 ай бұрын

AI is advance at light, what AI system or program do you use for your videos?

@Gemstudio-r1w 2 ай бұрын

Lumalabs

@pOOL_pANTS 2 ай бұрын

byson got that thanos look 🎱🩳

@hurtchie5949 2 ай бұрын

Soundtrack?

@AdeelMoti 2 ай бұрын

Why AI can't generate perfect fingers

@The24soldier 2 ай бұрын

Since Google sucks sometimes 😅 can someone explain to me how exactly A.I. generates these? Thank you, Strangers!

@nonconsensualopinion 2 ай бұрын

Neural nets have seen billions of images. They have been trained on them. So if shown an image, a neural net can give it a score of how real it is. Cat with 5 legs? Bad score. Nearly perfect looking cat? Good score. Now start with a completely random image, like tv static. Tell the neural net it's supposed to be a woman dressed as a ninja (remember, it has seen billions of pictures of women and many of ninjas). It of course gives a poor score. But, we can also see why the score was poor. From the mathematics the neural net uses to score the image, we can see how to slightly tweak each pixel of the noise to get a slightly better score. So, tweak those pixels. Check again. The score is still very poor, but ever so slightly improved. Keep tweaking, using the neural net score to guide you. It's like having a critic that tells you why they scored the image the way they did and you can use that feedback to improve. Eventually you have an image that is really good. That is one frame of your movie. Do the same for all frames of the movie. Now sprinkle in some techniques to the second frame looks like the same woman, and not just a different random woman. Some techniques to make sure the poses are sensible between frames, etc. These are basically the "diffusion" models, like "Midjourney". This is going to take a shit ton of computer processor power. On your PC it may take several seconds for a single image. 4k? Thousands of frames for a video? It's going to take server racks of power to get it done in a reasonable amount of time.