OpenAI Unveiled AI Video Generating Model Sora

Last week, OpenAI announced its latest development, Sora, a sophisticated text-to-video generating model capable of synthesizing high-quality videos.

This new model has yet to be available to the public. Still, it shows that OpenAI wants to dominate the AI video generation field, where other companies like Google are already competing with their tools. 

By loading the video, you agree to YouTube's privacy policy.
Learn more

Load video

Always unblock YouTube

OpenAI’s Sora: Precision in Text-to-Video Generation

According to the company’s announcement, Sora is a sophisticated model that can interpret long and complex prompts, synthesize high-quality videos, fill in missing frames, and more. 

It can create video scenes with multiple characters and different types of motion, including specific details for subjects and backgrounds. The software can process long and detailed prompts with considerable accuracy and can synthesize videos up to one minute long. 

Sora can create videos from a still image and fill in missing frames within a video sequence. The model's output videos shared by OpenAI have a fairly realistic feel. Still, a few details give away the AI nature of the clips, such as unnatural movement in animals, inaccurate facial features in human characters, or suspicious floor movement in indoor scenes. 

The developers acknowledge that Sora still has some challenges, particularly interpreting the physics of complex scenes and cause-and-effect concepts. However, Sora is an impressively good video generator. 

Red-Teaming for Safety and Creative Feedback

As we said earlier, the Sora model still needs to be deployed in any OpenAI product available to the public. The company says the unveiling of this model is now to give users a preview of what their generative AI will be capable of soon. If you ask me, they just wanted to keep up with Google and Microsoft, the other big firms racing into AI video generation, and video specialists who have been developing this technology for a while, like Runway and Pika. 

However, Sora is already accessible for selected “red teamers,” experts who are adversarially testing the model regarding misinformation, harmful content, and bias, among other significant concerns. The goal is to ensure the model is safe, particularly to overcome the issue of realistic deepfakes. 

Furthermore, the AI startup shared the same safeguards implemented in their Dall-E 3 model for image generation –which, among other things, reject potentially problematic prompts like those including famous names– and C2PA-compliant metadata will be integrated into Sora-based products and their outputs. 

Finally, the new AI video generator is also available for handpicked visual artists pursuing feedback from the creative community. 

It’s never been clearer that 2024 will be the year of AI video. 

Are you excited about trying OpenAI Sora? Share your thoughts!

THE AUTHOR

Ivanna Attie

All About Ivanna

I am an experienced author with expertise in digital communication, stock media, design, and creative tools. I have closely followed and reported on AI developments in this field since its early days. I have gained valuable industry insight through my work with leading digital media professionals since 2014.

• NETWORKING