A Preview of Sora, the New OpenAI Tool That Creates Remarkable AI-Generated Videos

A lit­tle over four years ago, we fea­tured here on Open Cul­ture a set of real­is­tic images of peo­ple who don’t actu­al­ly exist. They were, as we would now assume, whol­ly gen­er­at­ed by an arti­fi­cial-intel­li­gence sys­tem, but back in 2018, there were still those who doubt­ed that such a thing could be done with­out furtive human inter­ven­tion. Now, after the release of tools like Ope­nAI’s Chat­G­PT and DALL‑E, few such doubters remain. In recent weeks, anoth­er Ope­nAI prod­uct has caused quite a stir despite hav­ing yet to be prop­er­ly released: Sora, which can use text prompts to cre­ate not just replies in kind or still images, but minute-long video clips.

“This is simul­ta­ne­ous­ly real­ly impres­sive and real­ly fright­en­ing,” says Youtu­ber Mar­ques Brown­lee in his intro­duc­tion to Sora above. He exam­ines some of the demo videos released so far by Ope­nAI, high­light­ing both their strengths and weak­ness­es.

It would be dif­fi­cult not to feel at least a lit­tle aston­ish­ment at the result Sora has pro­duced from the fol­low­ing prompt: “A styl­ish woman walks down a Tokyo street filled with warm glow­ing neon and ani­mat­ed city sig­nage. She wears a black leather jack­et, a long red dress, and black boots, and car­ries a black purse. She wears sun­glass­es and red lip­stick. She walks con­fi­dent­ly and casu­al­ly. The street is damp and reflec­tive, cre­at­ing a mir­ror effect of the col­or­ful lights. Many pedes­tri­ans walk about.”

There’s some­thing Blade Run­ner going on here, in more sens­es than one.  The not-quite-human qual­i­ties about this “footage” do stand out on clos­er inspec­tion, and in any case make the whole thing feel, as Bown­lee puts it, “a lit­tle bit… off.” But as he also empha­sizes, repeat­ed­ly, it was just a year ago that the bizarre AI-gen­er­at­ed Will Smith eat­ing spaghet­ti made the social-media rounds as a rep­re­sen­ta­tion of the state of the art. The under­ly­ing tech­nol­o­gy has clear­ly come a long, long way since then, and though the mate­r­i­al so far released by Ope­nAI may feel faint­ly awk­ward and “video-gamey,” they clear­ly show Sora’s capa­bil­i­ty to cre­ate videos plau­si­ble at first and even sec­ond glance.

This may spell trou­ble not just for those cur­rent­ly in the stock-footage busi­ness, but also for those who hap­pen to believe every­thing they watch. Brown­lee calls the impli­ca­tions “insane­ly sketchy dur­ing an elec­tion year in the US,” but he may take some com­fort in the fact that Sora is not, at the moment, avail­able to the gen­er­al pub­lic. There are also explain­ers, like the one from the Wall Street Jour­nal video above, in which AI-indus­try pro­fes­sion­al Stephen Mess­er points out the tell­tale glitch­es of AI-gen­er­at­ed video, many of which have to do with the fin­er details of physics and anato­my. And if you find your­self pay­ing unusu­al atten­tion to the con­sis­ten­cy of the num­ber dig­its on Mess­ner’s hands, just tell your­self that this is how it feels to live in the future.

Relat­ed con­tent:

How Will AI Change the World?: A Cap­ti­vat­ing Ani­ma­tion Explores the Promise & Per­ils of Arti­fi­cial Intel­li­gence

Arti­fi­cial Intel­li­gence Cre­ates Real­is­tic Pho­tos of Peo­ple, None of Whom Actu­al­ly Exist

A New Course Teach­es You How to Tap the Pow­ers of Chat­G­PT and Put It to Work for You

Arti­fi­cial Intel­li­gence, Art & the Future of Cre­ativ­i­ty: Watch the Final Chap­ter of the “Every­thing is a Remix” Series

DALL‑E, the New AI Art Gen­er­a­tor, Is Now Open for Every­one to Use

How Peter Jack­son Used Arti­fi­cial Intel­li­gence to Restore the Video & Audio Fea­tured in The Bea­t­les: Get Back

Based in Seoul, Col­in Marshall writes and broad­casts on cities, lan­guage, and cul­ture. His projects include the Sub­stack newslet­ter Books on Cities, the book The State­less City: a Walk through 21st-Cen­tu­ry Los Ange­les and the video series The City in Cin­e­ma. Fol­low him on Twit­ter at @colinmarshall or on Face­book.