OpenAI Launches Sora, the Text-to-Video AI Model Redefining Digital Content

OpenAi, the company behind ChatGPT, has just revealed its new AI model called “Sora” that converts text to video. Sora is derived from the Japanese word meaning “Sky”.

Sora is capable of making video based on text prompts.

How Does Sora Work?

Sora is basically an AI model. It understands the purpose of the prompt and how the things within the prompt relate to the surrounding word.

It has been trained on possibly tens of thousands or maybe millions of videos. Responses are based on those clips.

OpenAI shown some examples as below;

“A movie trailer featuring the adventure of a 30- year old space man wearing a wool knitted motorcycle helmet”

Sora’s output is as below ;

Another example;

“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”

Below is the Sora’s output:

Both the generated videos seem quite accurate for the given prompts.

But, sometimes it can go wrong as shown in below videos;

“Archeologists discover a generic plastic chair in the desert, excavating and dusting it with great care.

Here, Sora failed to represent the chair as a rigid object resulting in accurate physical interactions.

Sora has basically merged different public videos in its library. It has videos of archaeologists. It has video of chairs. It also has videos of excavating. So, the output comes out as a mess-up. It has been learnt from those clips to generate output.

Let’s look at another prompt;

“Step-printing scene of a person running, cinematic film shot in 35mm” , and look at the output below;

It shows a person running backward on treadmill, which is clearly impossible. This is a flip side of the advanced technology.

Sora is OpenAI’s advanced technology. There are numerous uses of it; you can make stock footage from it, you can use it for storytelling, video campaigning, even in news story. All these make Sora a very consequential tool.

Sora’s Availability and Usage Restrictions

It’s currently not available to the public. The company is providing access to some creators and researchers to get their input on how to improve the model so that it best serves the needs of creative professions. The generated video is restricted to one minute. It will not create sexual content, violence, or celebrity likeness.

We have gone pass by funny chatbot. This is now a bending reality. People may dismiss fake news claims on social media but what if those claims back by pictures and videos. It could amplify those fake news.

The restrictions are not enough. There are many unregulated models out there. Those willing to bend the rules. We have already seen examples in past like fake robot call by US president Joe Biden, or fake porn images of Taylor Shift.

Tech companies claim to have solutions in place. Watermark is among the items. Watermarks are used on all photos and videos to make them easily identifiable. They are in the Sora videos too. It’s a temporary solution until new regulations are implemented.

Where does AI go next?

Technology is constantly improving, whether you like it or not. Soon these videos will become indistinguishable from the real ones. We need regulations for when that happens. In the era of artificial intelligence, companies by themselves are unable to ensure safety. The government needs to take the lead and enforce the rules, whether it is a digital signature or a watermark.

ChatGPT has launched in 2022 and now the market is filled with text-based AI models. There are many others who are doing the same thing. Therefore, strong laws and regulations are urgently needed.