OpenAI today introduced its text-to-video creation tool called “Sora”. The model, which has now been tested by security experts, can create frighteningly realistic 60-second videos.
OpenAI, the developer of ChatGPT, has added a new one to the artificial intelligence models that are taking the world by storm today. “soraThe model called ” can create video from text. The sample images shared by the company show that the model is frighteningly good.
Sora is currently being tested by experts on issues such as disinformation, hate content and bias. The company states that it wants to take the necessary security steps before putting the model into use.
Sora can create realistic 60-second videos with text commands
“The beautiful, snowy and vibrant city of Tokyo. The camera moves down the busy street of the city, following several people enjoying the beautiful snowy weather and shopping at nearby stalls. Magnificent sakura petals are flying in the wind with snowflakes.”
*If you can’t view the video, click the link here.
As we mentioned at the beginning, Sora is a text-to-video creation model. In other words, it can convert a text command you type into a video. OpenAI, Sora 60 seconds, He explained that he can create videos of up to one minute.
of these videos visual quality remains high, is highly detailed, includes complex movements, and has multiple characters with vivid emotions It was stated that he could accommodate. When we look at the videos, we can see that it is frighteningly realistic. OpenAI’s statement is as follows:
“Sora can create complex scenes featuring multiple characters, specific movements, and accurate details of the plot and background. “The model understands not only what the user wants in the command, but also how those things exist in the physical world.”
Sora will be opened to people like artists, filmmakers, and designers, as well as testers for security. These people will provide feedback using the artificial intelligence tool and contribute to the further development of the model.
It can also produce videos from images
*If you can’t view the video, click the link here.
Apart from this, OpenAI added that it can create videos not only from text but also from images. According to the company’s statement, it can take an existing image and produce a video from it. In this video, too, he pays attention to small details and details.
It is not without some weaknesses
Of course, since the model is still in the testing stages, it also has weaknesses. OpenAI also acknowledges this. The company sometimes analyzes the physics of complex scenes. may have difficulty simulating it accurately. And may not understand cause-effect examples says.
For example, in a video, we can see that a person bites a cookie, but afterwards there is no bite mark on this cookie. It is also said that it may confuse details such as left and right and may have difficulty explaining events that may occur over time, such as following the movement of the camera. You can see examples of these errors above.
RELATED NEWS
OpenAI Allegedly Developed a Search Engine to Compete with Google
However, when we look at the sample videos in general, we can see what great potential Sora has. It is obvious that it has the potential to revolutionize artificial intelligence models. We’ll see what he can do when he comes to everyone.
Other video examples produced with Sora
*If you cannot view the video, click the link here.
*If you cannot view the video, click the link here.
RELATED NEWS