This AI takes videos to a whole new level – and the result is pretty scary

‘Sora’ sounds innocent, but is currently the most advanced video generator. It can create almost photorealistic videos of people, animals or landscapes with complex camera movements from simple text instructions.

For example, it looks like this:

According to the information, the published sample videos are completely created by artificial intelligence. In the video above with the woman, the text instructions were for her to wear a leather jacket and a red dress and walk confidently down a street in Tokyo. The street is wet and reflective, creating a mirror effect from the colorful neon lights.

The AI ​​model behind Sora was developed by OpenAI, the Microsoft co-funded technology company that developed the chatbot ChatGPT and the AI ​​image generator DALL-E. The latter is a computer program that creates images from text descriptions based on machine learning. Now OpenAI is taking it one step further with AI-generated videos.

Fascinating and extremely dangerous

Sora can also modify existing videos or generate short clips from images. None of this is new, but it is at a level never seen before. Most people therefore react with a mixture of fascination and fear. This is absolutely correct. The latest launch of OpenAI is impressive, the potential is enormous in both positive and negative ways.

The fears often expressed by critics include simplified election manipulation through deepfakes of politicians, AI-generated revenge porn at the touch of a button, and a flood of copyright infringements. In short: Sora is extremely dangerous. Because history shows us that all technology can and will be misused for evil.

OpenAI is aware of the dangers. Watermarks are intended to identify AI videos. The company wants to prevent revenge porn by not accepting requests that require extreme violence, sexual content or resemblance to real people. Moreover, the public, apart from carefully selected people, has not yet been granted access. Experts should first explore possible safety risks before publishing the program.

So it’s only a matter of time before Sora is released. OpenAI wants to make money with it in the foreseeable future. This is also an understandable request, after all, the vast majority of customers will probably use the AI ​​generator honestly.

To help convey Sora’s capabilities, OpenAI has released dozens of sample videos. An overview.

Sora makes a movie trailer

The quality of the AI ​​videos also silences famous YouTubers. In response, MrBeast, the world’s most successful YouTuber, who is known for his elaborate videos, wrote to Sam Altman, CEO of OpenAI: “Sam, please don’t make me homeless.” Altman replied: “Can you make a video, what do you want?”

Sora won’t disrupt the film industry in the coming months with these 60-second clips, but video platforms like YouTube and TikTok could be in for a big change. AI should enable many people to produce high-quality content for social platforms.

Technical measures are intended to prevent users from, for example, staging historical-looking AI videos as real recordings. However, experience shows that people with sufficient criminal energy almost always find a way to circumvent such barriers.

Sora also competes with drone videos

Nevertheless: Generative AI is making rapid progress, as this comparison with an AI video from early 2023 shows.

Sora isn’t perfect, but it makes other text-to-video generators look pretty old

The same instruction with Sora and the competition Stable Video

Sora can also modify existing videos. Here a car drives through a mountainous forest landscape

You can ask Sora to keep the original video, but move the action to winter

Or: “Convert the video to a pixel art style.”

Last but not least, Sora can bring images and photos to life

These videos and many others on the OpenAI website show: The genie is out of the bottle. And Sora is just the tip of the spear, more AI tools will follow. It is up to people to create something creative with it.