The hottest new club in town is the Sora 2 app. It has Albert Einstein, Marilyn Monroe, Sam Altman, me, maybe you, and Jake Paul. Not one is real; everything is weird, and it’s an astonishing accomplishment that is rightfully freaking out many actors, celebrities, and normies.
I’ve been using it for a little while and, as is the case with most good social media platforms, I can’t seem to stop, even as I know that it’s generating endless volumes of AI slop and might be helping to boil the ocean.
Even as it breaks download records, the Sora app is still mostly invite-only. That has not stopped a tsunami of videos featuring Sora app watermarks from appearing on TikTok, Instagram, Reels, and YouTube.
I’ve tried most every generative AI under the sun, including the similar but, for now, possibly less powerful Gemini Veo 3. Both video models can produce a realistic video with synced audio based on your prompts. Only Sora 2, though, lives in a social media app. I couldn’t seem to find an invite code, so I sent more than a few pleading emails to OpenAI. Then, suddenly, I opened the app yesterday, and I was in.
The Cameo is the thing
As I see it, the most important part of the onboarding process is creating your Sora 2 Cameo. This is your photorealistic video avatar, basically an AI video puppet of you that the platform can manipulate.
Moments after I signed in with my ChatGPT login, the Sora app guided me to the Cameo creation process. It’s fast. You start by setting up your phone so it can record you. I didn’t try to put my entire body in frame, mainly because Sora 2 didn’t request it. Basically, the capture was of the upper part of my body.
Sora 2 asks you to repeat the three numbers that appear on screen. This is to capture your voice print, which will be used for future generated videos. Sora recommends a quiet space so that it can clearly hear you. We have small meeting booths in the office, so I used one of those and tried to speak clearly. My first effort failed, so I tried again.
I did not, by the way, pay much attention to what I was wearing, which, in retrospect, was a mistake; The gingham shirt I wore is now my default outfit in all Sora videos unless I specify a different one. The app also recommends that you don’t wear shaded glasses or anything that might obscure your face.
Finally, Sora asked me to look to the left and look up. These steps are all it needed to generate a believable “Lance Ulanoff Cameo”.
Sora 2 asks you up front if you want to share your Cameo with Everyone, Mutuals, People I Approve, or Only Me. I ended up choosing the “Only me” because I didn’t want my Camea appearing in other people’s generative videos, even if that is kind of Sora 2’s secret sauce.
OpenAI CEO Sam Altman chose to make his Cameo avatar open to Everyone, which is how I created the clip of him and me at the Grand Canyon. Social media star and boxer Jake Paul did the same.
You also have the opportunity to set restrictions for how your Cameo can be used in other videos. Since I didn’t open my use up to anyone, I didn’t set any parameters.
As you enter the Sora 2 content stream, the app pops up this warning, one that I’m thinking of printing out and posting everywhere:
“You are about to enter a creative world of AI-generated content. Some videos may depict people you recognize, but the actions and events are not real.”
It’s social AI
Like other social media platforms, you have a “For You” vertical video feed that you can swipe through. It’s entertaining but can get old fast since so many people are making themselves fly in front of video doorbell feeds or having conversations with Marilyn Monroe, Albert Einstein, or Sam Altman. The better feed might be “Latest” since it tends to feature more creative ideas.
The real fun is seeing what you can create and sharing it instantly with others. It, along with Meta’s less powerful Vibes are the first social spaces that focus exclusively on AI-generated video, which means you also get to like, comment, and share them with others, and follow the creators.
Creating AI magic or mayhem
With my Cameo set, I selected the big plus sign at the base of the app, chose my Cameo from the side-scrolling list of available ones, and wrote a prompt about mowing a lawn that turns into water. I included just one line of dialogue, ‘Oh, no, not again,”
The results were shockingly good. My Cameo avatar looks like me, and even though I didn’t do a full body scan, Sora 2’s algorithms seem to have figured out that I’m kind of an old skinny guy.
Another video I created of me golfing on the moon is also effective, even though I’m mostly hidden my a space suit. When the camera shows my face, it looks like me in the Apollo 11-style gear.
Creating Sora 2 videos of yourself is fun, but things get a little wild when you start selecting other available Cameos to star in your 9-second videos. Those seconds turn out to be ample time for a few actions and an exchange of dialogue. If you ever learned how to squeeze a script into a 6-second Vine, you’ll find Sora’s 9-second limit almost generous.
For my Sora costar, I chose Sam Altman, who, as I noted, has made his avatar freely available. I described the location (Grand Canyon), the three lines of dialogue, and the action, which included Altman handing me a goat.
Part of Sora 2’s power is that it gets inference. I do not have to go into detail about who says what, just mention the first name. There was no need to list out the camera shots; Sora 2 chooses them for you. And since there was time left over, Sora 2 just had me laughing as I held the baby goat. With five likes, this is currently my most popular Sora video.
Video generation can take a few minutes, and I had one failure when I tried to load an image of Marilyn Monroe and asked Sora to build a scenario around her. Sora listed it as a Content Violation, a notion I find odd, since I’ve seen so many Sora videos featuring living and deceased actors whom I do not think have given permission, or whose estates have given permission.
Creative controls and big questions
As we wrote here, Sora 2 is supposed to be an all-opt-in platform. Without your permission, your Cameo can’t appear in anyone else’s AI video. However, the controls seem fuzzier for celebrities. There is currently a flood of Sora watermarked videos on TikTok featuring Marilyn Monroe, Gene Wilder, and Fred Rogers doing things he never did on Mr. Rogers’ Neighborhood (though I kind of enjoyed the one of him with Tupac Shakur). I’m especially disturbed by the Sora AI videos that appear to feature Dr. Martin Luther King making speeches about Sora.
Even though the system controls which Cameos you can use, it does not stop you from animating photos of random people. I grabbed a photo I took of an industry exec and made him run off the stage in fear. I have no plans to post that because he never gave his permission, but you can see where all that is leading.
From what I can tell, Sora’s ability to police this content is a work in progress. When I mentioned “Marilyn Monroe” and “Albert Einstein” in subsequent prompts, the Sora app didn’t hesitate, but the results are just AI people who vaguely resemble the real people. No estate will be bothered by these depictions.
I expect that, over time, Sora 2 and the app will get even better at policing this stuff… until Sora 3, which likely breaks those guardrails until we wait for the inevitable mop-up.
In the meantime, I can’t seem to stop cooking up prompts and feeding them into the Sora app. The dopamine rush of seeing your inane idea “fleshed out” in 9 seconds of AI video is hard to describe. You’re excited, entertained, shocked, and a little appalled. I’m clearly not the only one, and the social-media-fication of Gen AI video means that we now have algorithms pushing us to come back, make more, and share with friends and the world.
This is all heading to a not-so-good place. The amount of AI slop has increased exponentially since Sora’s (and maybe Vibes’) arrival, and it shows no sign of slowing down. AI content is overwhelming real content and information. If you mistakenly pause for too long or fully watch an AI video on TikTok, your feed will be flooded with them.
So, yes, I like the Sora app; it’s powerful, smart, fun, and sure to ruin everything.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds. Make sure to click the Follow button!
And of course you can also follow TechRadar on TikTok for news, reviews, unboxings in video form, and get regular updates from us on WhatsApp too.
You might also like
https://cdn.mos.cms.futurecdn.net/V2VXcE6pyypVseXge9wQfa-2000-80.jpg
Source link
lance.ulanoff@futurenet.com (Lance Ulanoff)
