OpenAI Unveils Sora: Text-to-Video AI Model

MetaversePlanet February 16, 2024

0 2 minutes read

OpenAI recently introduced a text-to-video creation tool named “Sora.” Currently under evaluation by security experts, this model is capable of producing alarmingly realistic videos up to 60 seconds in length.

OpenAI, the organization behind ChatGPT, has launched another groundbreaking artificial intelligence model, “Sora,” which generates videos from text descriptions. The sample visuals released by the company demonstrate the model’s remarkably high-quality output.

“Sora” is undergoing expert evaluation on issues such as misinformation, hate speech, and bias. The company has expressed its intention to implement necessary security measures before making the model publicly available.

Contents show

Sora can create realistic 60-second videos with text commands

OpenAI Unveils Sora: Text-to-Video AI Model

As previously mentioned, Sora is a model capable of creating videos from textual descriptions. This means it can transform a typed text command into a video. OpenAI has detailed that Sora can generate videos lasting up to 60 seconds, or one minute.

The company has indicated that the visual quality of these videos can be maintained at a high level; they are exceptionally detailed, encompass intricate movements, and can portray multiple characters exhibiting vivid emotions. The videos produced are alarmingly realistic. OpenAI’s description of Sora’s capabilities is as follows:

“Sora is capable of crafting complex scenes featuring multiple characters, precise movements, a narrative, and accurate background details. The model not only grasps the user’s instructions but also understands the real-world existence of these elements.”

Sora will be made available to individuals such as artists, filmmakers, and designers, in addition to safety testers. These users will employ the AI tool to offer feedback and help refine the model further.

It can also produce videos from images

Furthermore, OpenAI has announced that it can generate videos not only from text but also from images. According to the company, it is capable of taking an existing image and creating a video based on it, paying close attention to small and intricate details.

It is not without its weaknesses

Indeed, since the model is still undergoing testing, it has its limitations, a point on which OpenAI concurs. The company acknowledges that the model can sometimes struggle to accurately simulate the physics of complex scenes and may fail to grasp cause-and-effect relationships.

For instance, in a generated video, we might observe a person biting into a cookie, yet the cookie remains unmarked. It’s also noted that the model can confuse details such as left and right and may find it challenging to depict events that unfold over time, such as tracking the camera’s movement. Examples of these errors are provided for illustration.

Nonetheless, a review of the sample videos overall reveals the immense potential of Sora. It’s clear that it holds the promise to bring revolutionary changes to artificial intelligence models. The full extent of its capabilities will be better understood once it becomes widely accessible.

*If you can’t view the video link here Click.