Open AI's latest wonder: text to video with Sora

Don’t worry, the fully-equivalent, completely open-source Chinese version will be here before we know it.

Miyazaki is not pleased:
Imgur

Somebody managed to do the cat:

I have no idea what Studio Ghibli is, but these pictures are pretty hillarious

Blink.

Blink blink.

(Stares in incredulity.)

It’s just so good:
Imgur
Imgur
Imgur
Imgur
Imgur
Imgur
Imgur

It also can Lego. (Look far down this page)

Color me ignorant of it as well.

Any fan of animation knows the name Studio Ghibli. The Japanese animation studio has made a name for itself by crafting some of the greatest animated films of all time. The studio was founded in 1985 by established animators Hayao Miyazaki and Isao Takahata and producer Toshio Suzuki. Over the next three decades and 23 movies, Studio Ghibli produced some of the best coming-of-age, fantasy and family films of all time. Studio Ghibli is a creative and box office powerhouse. Five of the studio’s films are among the highest-grossing anime feature films and five have received Academy Award nominations in the states.

Well, I’m not into anime or manga, so I guess that explains it. Been playing around with it on Chat GPT for the last half hour, and it’s pretty impressive (still fucks up hands sometimes.) How long until a similar technology comes into Photoshop and lets you fix shitty lighting and stuff like that. What’s funny is that I fed a photo of myself recently winning first place at a chili cookoff, and it changed my awards certificate first to “dad coach of the year” and second time to “first place chess club tournament.” Lol. But I’m impressed at it filling in sensible English words.

Still boggles my mind. It is completely on the level of having never heard of Disney.

Maybe to you. My gen Alpha kids, one of whom is into manga and vocoloid stuff had no clue what it was either. I asked my wife and a shrug from her as well. Not all of us live in the same sphere.

Speaking of Sora being aggressive about not violating copyright does anyone know who has the copyright on what Sora produces? You? The people who own Sora?

Same with ChatGPT writing an article for you. Who owns it?

(I tried a search…seems like something that would have been discussed here already…I think it has…but I could not find it.)

Whatever the secret sauce for getting the good renders is, I don’t seem to have it. I tried using the ChatGPT app instead of the web interface. Uploaded this Joro photo of mine.

Asked for Ghibli style, it refused. Asked for Lego, it made something, but nothing like a transformation of my photo.

Asked for The Simpsons, got this.

Uploaded a kitten photo

Got this for Simpsons style

And then after generating only three photos won’t generate another one for 24 hours. But from three “successful” images, none of them were the magic transformations. Must be giving me older versions of ChatGPT or the image generator.

This is the best individual thread I have found of outputs

Interestingly 4o still fails at showing a watch dial with a particular time, a well-known glitch in AI image-generation, which have been trained on photos with a time of 10:10.

And South Park too:

Impressive, although trump seems to have four fingers in one hand and five in the other.

My initial prompt was simply “turn this photo into studio ghibli style” and subsequent prompts were “now try this one!” “Do this one!” Etc. I’m using GPT 4o and I do have the $20/mo subscription. It follows the source material pretty closely in terms of composition and coloration. And I haven’t run into being locked out of image generation yet, but I’ve only done about 7 of them so far, as the initial novelty and amazement is kinda of wearing out already for me. I’m quickly jaded it seems. But I will use it again, I’m sure.

ETA: Building on the South Park style, I tried Simpsons and that worked well, too.

His secret is out!

Sorry, not everyone watches anime, I guess?

I liked Pokemon when I was a kid (when it was the OG roster), but I haven’t really watched much since.