The moment that many feared has arrived: AI is now capable of creating realistic videos, very difficult to distinguish from real videos. Sora's announcement earlier this year was just a small taste of what's possible, and to everyone's surprise, OpenAI decided to ignore the criticism and artists' fears and go public with the new AI.
Maybe the launch of Sora is a counterattack against Googlewhich last week launched a video-generating AI called Veo; However, Veo is available on a very limited basis, only to business users selected by the company. Instead, OpenAI has been much more ambitious with the launch of Sora.
YouTuber Marques Brownlee was responsible for debuting Sora in a presentation video, but now the AI is available to all users, albeit with limitations; For example, In Europe and other countries it is not yet possible to use Soraand it's unclear if this is due to stricter laws in the European Union or if it's a way to limit the impact on the company's servers.
And it's the most demanding AI yet, capable of taking several minutes to create the video requested by the user; Perhaps this is why Sora was released independently and is not integrated into ChatGPT, at least for now. Sora is free, but there is an exclusive Turbo version for ChatGPT Plus and Pro users.
What you can do now is go to Sora's website and see some of his creations; and there is no doubt that they are surprising. Although the usual problems with video-generating AIs continue to be noticed, such as unnatural movements and elements coming out of nowhere, the level reached by this technology in just a few months remains shocking. Limbs remain the weak point of OpenAI's AIs, and Sora confuses legs and arms in animals and humans.
This can be REALLY good for landscapes. Almost any drone shot of a prominent landmark could pass for stock footage or is VERY close to being usable for a shot in a documentary or low-budget film. pic.twitter.com/nY4khuVP4T
— Marques Brownlee (@MKBHD) December 9, 2024
Despite everything, the image quality is very good, with resolution videos up to 1080pbut these are the ones that take the longest to generate; The fastest option is 480p videos, which take up to 8 times less time, while the middle option is 720p resolution, which takes 4 times less time.
Sora has several modes, depending on what the user wants to create. Like ChatGPT, it is also capable of receiving a command, or “prompt”, and creating a video from scratch that matches the description. But what is perhaps more interesting is the ability to create videos from an image uploaded by the user, in addition to editing videos already created to adapt them to our needs.
The other features, like “remxing” videos or turning images into videos, can be useful tools if you know what you're doing – but the most consistent conclusion for me is that the models don't know which direction or what speed makes sense. objects in this specific image. SO… pic.twitter.com/1Y7g4mjFIY
— Marques Brownlee (@MKBHD) December 9, 2024
Related to the latter, the Re-Mix function is capable of creating new videos from existing videos, based on the description provided by the user; It is also possible to indicate the “strength” of the AI, that is, to what extent it should modify the original video, if we want something new but not too different. Another striking feature is the “Storyboard”, which allows join multiple video clips to create a story; each video will take into account the previous result to ensure a certain consistency, even if this remains the weak point of generative AI.
Sora is OpenAI's most controversial launch so far, due to the potential it has to end creative jobs. Remember that this AI was disclosed in advance as part of a demonstration; responsible users They accused the company of using the artists to improve his public image and as a “hook” to sell Sora to content creators, while ignoring their concerns and criticism.
This video contains a bunch of truncated text, telltale signs of AI-generated videos. But the cutaways, the moving text ticker, the newsreel-style shots… these were all things SORA decided to do themselves, and these news anchors looked very… real. pic.twitter.com/7iM3UQTC65
— Marques Brownlee (@MKBHD) December 9, 2024
Another aspect that already raises concerns is the ability of AI to create fake videos for advertising or spreading fake news. One of the videos Brownlee created demonstrates this possibility, showing what looks like a news video about a news story that didn't happen, with surprising quality. Sora still fails on aspects such as the words that appear in the video, but by changing this it would be easy to create a video of a news program telling fake news about the assassination of the president of a government, For example.
OpenAI's strategy in the face of these controversies is not to mention them. Although it initially boasted of collaborating with filmmakers and artists, this time around the company did not share or mention their creations, and avoided talking about professional use of Sora; instead, he said he hoped Sora would “help explore new forms of creativity.”
And about fake news, it says that all videos come with metadata that identifies them as being created with AI, but the company itself admits that it is “imperfect”; He also confirmed that he does not yet have a public tool to check if a video was created with Sora, leaving it up to the internet user to decide whether the video they are watching is fake or not.



