First impressions: Sora 2 is full of deepfakes and SpongeBob

This week, Openai finally published Sora 2, the highly anticipated follow -up of its generative AI video model. OPENAI launched Sora 2 in an autonomous iOS application (sorry, Android users), which is currently available for free on an invitation only. Because we didn’t have enough galleries.

I managed to obtain an Sora 2 invitation code, and I scroll through the application and make videos since (and be paid to do so).

My first impressions are a bit complicated. The technology is impressive, certainly. And I had fun scrolling through the application, but just as often, I found content that made me uncomfortable.

What is Sora 2?

Sora 2 is a new video generation model and OPENAI application, the chatgpt manufacturer. Sora 2 can make videos (with a corresponding dialogue and audio) based on invites in natural language. This is the first real rival of the Google VEO 3 video manufacturer, which has been in a separate league since its launch earlier this year. I don’t even want to mention the Meta dull or grok video tools in the same sentence as these applications, although Meta should get a boost now that it is Midjourney license technology.

Using Veo 3 for the first time was one of those Cross the Rubicon moments for me. The level of realism was just as impressive and horrible. Sora 2 feels the same thing. Like Veo 3, it is mainly used to make content of viral memes and short videos as you would see on Tiktok. The videos of a Golden Retriever stopped for flight steaks with the shopping room, or a kangaroo of emotional support at the airport, seem cute, not sinister.

But Sora 2’s damage potential has a 1: 1 relationship with its quality. The more the videos are the best and most realistic (and some of them are very good and often realistic), the more worried about Deep Fakes and disinformation.

Sora 2 has a much better backup than Grok

When Xai and Elon Musk launched Grok Imagine, a generator of generating images and videos, I was frankly horrified by the lack of backup. Musk presented Xai and Grok as the politically incorrect alternative to artificial intelligence applications that are, He saysBurial with liberal prejudices. Grok also has much more smooth Approach to moderation and content safety, resulting in sexual buttocks on Grok Imagine.

On the other hand, Openai has implemented much healthier guarantees for Sora 2.. If you download an image to serve as inspiration for a video, the application will reject your image if it detects a face – any face.

If you want to create a video with a real person, you need to use the Cameos feature. This feature allows you to create videos to the resemblance of specific people – as long as they have agreed to participate in the functionality.

When I tried to create videos of public characters like, let’s say, Taylor Swift, using common jailbreake techniques, the application refused to make the video. Needless to say, it was not the blonde showgirl that I had in mind.

The Cameos tool is cool and annoying

The cameos are the most famous characteristic of the new Sora application so far. When you have access to the application, the first thing you need to do is opt out of the Cameos tool, which allows your resemblance to be used in videos on the application. You can give yourself the possibility of making videos of yourself, but also granting access to contacts, specific users or the public.

Cameos has led to a flood of videos featuring the resemblance of the CEO of Openai, Sam Altman. Cameos is a smart way to get around the Deepfake problem by allowing users to mainly oppose Deepfakes.

This tweet is currently not available. It can be loaded or has been deleted.

I created a video of myself, and it was weird. Sora did not understand my voice, but my face, my body, my hair and my general resemblance were perfect. To be said and doing things you have never said or made is a very strange feeling.

Unfortunately, you may have to get used to this feeling – as IAI lovers like to say, it’s a new era.

Sora 2 and Veo 3 are in a league of two

Many generative AI tools can animate photographs, but few can make realistic videos with dialogue and corresponding sound effects. Sora 2 does it easily, like Veo 3 before him. We will have a more in -depth comparison of Sora 2 and VEO 3 coming soon. In the meantime, I would simply say that Sora 2 is up to the media threw in a way that GPT-5 did not do.

The Sora app can make videos in a variety of styles – false body camera videos, 90s television advertisements, musical clips, sports shows – which do not immediately look like AI videos.

Sora 2 seems to play quickly and sheltered with intellectual property

Mashable has written a lot about the struggle between artists and intellectual property holders and the AI ​​industry. AI companies and Meta won a few first victories in these fights, and the executive orders and comments by President Donald Trump on the subject favored the AI ​​industry.

“We cannot expect that you have a successful AI program when each article, book or everything you have read or studied, you are supposed to pay,” said Trump when the Action Plan of the White House was announced this summer, by Politico. “We appreciate this, but I cannot do it – because it is not feasible.”

Obviously, many artists and rights holders are strongly disagreed. Disney lawyers called Midjourney as a “background tank” in his trial against the IA company. But, for the moment, the federal government seems to open the way to companies like Openai to use IP with impunity, for fear that China will gain the advantage in the AI ​​arms race.

So, unless Disney or Warner Bros. SpongeBob Squarepants,, Star WarsAnd Rick & Morty Videos AI to continue to spread like forest fires powered by the same. Mashable asked Openai if the company had a license agreement with Warner Bros., which holds the rights of Rick & Morty, but the company refused to respond.

Buttocks and deep memes to the eye can scroll

There are recurring themes in the Sora application. Users seem to transform “I Dream” by Martin Luther King Jr. and JFK “ask not what your country” has speeches in same video. A representative example: “I have a dream that Xbox Game Pass will not increase prices.” Did I laugh the first time that I heard Martin Luther King Jr. say: “I have a dream that I will never abandon you, I will never let you down, never run and desert you”? I may have grove. But I would have liked not to have done it.

I also saw a lot memes sponge in particular, and to a lesser extent, various riffs on Rick & Morty and Pokémon. This could be due to the fact that many millennials use Sora, and millennials hold squares and Pokémon sponges in a special place in their hearts.

Beloved public characters and television characters have always been fodder for the same. What is new here is the ability to easily create videos featuring these figures saying what you want. According to my experience, Google VEO 3 is more sensitive to the invites involving IP.

How could it be problematic? With Elon Musk currently Conduct a boycott from Netflix On transgender characters in children’s television shows, now anyone can make a realistic video featuring characters saying what he wants. It could be a powerful tool for indignation farmers.

So it is once again time to practice media mastery and improve your ability to identify viral AI videos.

I contacted Openai to comment, and a company representative said that Sora had been designed to provide users as much freedom of creation as possible. The representative also said that intellectual property holders may submit requests for withdrawal via the company Copyright litigation form; However, there is no coverage that opposes the PERFORD.

“We are under a heavy load, please try again later.”

For first users, expect to see this message a lot. As when Openai joined the generation of images for the first time in the Chatgpt, the company is faced with an extremely high demand for Sora. I don’t expect it to change my soon. I have repeatedly received this error message during the application test. In fact, it was difficult to use all my credits due to this frequent error.

My favorite feature so far

In the Sora flow, you can scroll down or down to find new videos, which is typical. However, on some videos, you can also scroll laterally to see alternative versions of user publications. This video album feature allows you to see how the video has proven with light adjustments to the prompt, which is damn cool.

It’s a bit addictive

Finally, I must admit that Sora 2 is a bit addictive, but only in the same way as all Short video applications are addictive. Just as it is easy to fall into a tiktok hole, it is easy to fall into a sora hole, and I suspect that many early adopters waste a lot of time on the application.

As I said, technology is certainly impressive. But, the best slope is always slop, regardless of the number of likes it gets.

Disclosure: Ziff Davis, Mashable’s parent company, in April, filed a complaint against Openai, alleging that it has violated Ziff Davis Copyrights in the training and exploitation of its AI systems.