OpenAI Launches Text-Based Video Creation Tool

According to CNBC , OpenAI has just introduced Sora - a next-generation AI model that works similarly to their previous DALL-E image-generating AI tool. Users simply input the desired scene and Sora will return a high-resolution video clip. Sora can also create video clips inspired by still images and expand existing videos or fill in missing frames.

Samoyed and Golden Retriever dogs frolic around a city with neon lights at night in AI-generated video

Video creation could be the next frontier for creative AI, after chatbots and image generators made inroads into the consumer and business worlds not long ago. While content creation will excite AI enthusiasts, these new technologies are raising serious concerns about misinformation as major global political elections approach. According to data from machine learning company Clarity, the number of AI-generated deepfakes has increased by 900% year-over-year.

With Sora, OpenAI is looking to compete with video-generating AI tools from companies like Google and Meta. Similar AI tools are available from other startups, such as Stability AI, which has a product called Stable Video Diffusion. Amazon has also launched Create with Alexa, a model that specializes in generating short-form animated content for kids based on commands.

Sora is currently limited to generating videos a minute or less long. OpenAI has made multimodality, a method that combines text, image, and video generation, a goal in an effort to provide a broader set of AI models.

So far, Sora has only been made available to a small group of safety testers, or “red teams,” who test the model for vulnerabilities in areas like misinformation. The company has not released any public demos beyond the 10 sample clips available on its website, and says accompanying technical documentation will be released soon.

OpenAI also says it's building a "detection classifier" that can identify Sora-generated video clips, and it plans to include certain metadata in the output to help identify the content created by the AI. That's the kind of metadata Meta is looking to use to identify AI-generated images.

Sora is an AI model that uses the Transformer architecture that Google researchers introduced in a paper in 2017. In its announcement, OpenAI said Sora serves as a foundation for models that understand and simulate the real world.

Source link: https://thanhnien.vn/openai-ra-mat-cong-cu-sang-tao-video-bang-van-ban-185240216160759311.htm