Bloopers & News 05/02/24

My first attempt at AI videos.

People are waiting with bated breath for the release of Sora, OpenAI’s generative new video software. Here is the ad to make you understand why:

You can also scroll through all the individual videos here: https://openai.com/index/sora

To me the most interesting clips are the images that I’ve never seen before: a boy sitting on a cloud reading a book, a coral reef made of folded paper, and pirate ships battling within a cup of coffee. The moments of a car driving down the road or a stack of TVs are less compelling, but I assume Sora wants companies to drool over the advertising potental (hey Toyota, want to advertise your car without renting a helicopter to film it?)

Creators’ personal and crazy ideas are always more engrossing than the images we already know, and that is what should comfort artists. No matter how powerful the software it can only be as interesting as the prompts it is given.

The job of “prompt creator” is going to be huge. Remember in the early naughts when people started calling themselves “social media experts” and we thought they were dopes? That will happen with prompt creators. It will seem like the dumbest job ever and then suddenly they will be the key employee at your next company meeting. There will surely be some clever new name. Prompt artist?Prompt auteur? (Promp-teur?)

Curation will go hand in hand with the Promp-teurs. AI will be pumping out SO MUCH TRASH that it will be someone’s job to separate the chaff from the wheat. And for several years this curator will actually need a background in photography,art history, graphic design, or advertising to select the good stuff from the mediocre. Graphic designers will continue to clean up images and fix the AI “hallucinations” (which is what people are now officially calling the mistakes that AI makes, like giving a person three legs).

For AI video creation, I suspect cinematographers and directors will be still be necessary for some time to direct the “camera work” of these videos. You will see in the Sora ad that the prompts included camera instructions like “close-up” and “tiltshift.” Eventually, AI will begin to intuit when the shots should be establishing, long, sun-lit, etc. and then I suppose filmmakers will also become curators.

But don’t panic just yet. I (of course!) decided to play around with some AI video creation, which are this week’s NAILED IT.

This first one was created with an app called InVideo, which is what is offered on OpenAi (I suppose until Sora is out). My prompt was “in graphic novel style we follow a woman with long red hair and glasses walking down a street in Barcelone, Spain. She looks behind her, worried she is being followed.”

The first version it generated was all live action so I repeated, “please make it look animated in the style of a graphic novel.” (Yes, I say “please” to the AI because I am a polite southern girl who can’t help herself.)

The opening and closing images are sort of cartoony drawings, but all the other footage seems to come straight from stock video footage and is even watermarked “iStock.” This is FAR from generating original content.

I then tried a program called AppyPie which describes itself as “AI Text To Animation.” It is free for 7 days and then 90$ a year. I gave it the same prompt as above: “in graphic novel style we follow a woman with long red hair and glasses walking down a street in Barcelone, Spain. She looks behind her, worried she is being followed.”

This looks more like a graphic novel and seems to have actually been generated by the AI, as opposed to it scraping clips from the internet. However, it has no character consistency, and the 2 second clip took around 20 minutes to produce! So no need to worry that people will be pumping out full length features anytime soon.

As for bloopers from the graphic novel this week, I’m shocked to report that I only had one and it is still pretty darn good. The prompt was “A 45-year-old red haired woman with glasses sitting talking at a kitchen table with a woman with long black hair.

The women are not talking to one another and for some bizarre reason the red head is declaring her age, which is what we middle-aged woman love to do all day long.

---