Arrr, mateys! Be holdin' OpenAI's latest treasure, the Sora text-to-video model! It conjures unbelievably real content, ye scallywags!
2024-02-15
Yarr, Sora be havin' th' power t' craft clips o' lifelike scallywags, plunder, 'n critters, but alas! It be strugglin' wit' certain parts, ye see?
OpenAI has announced its first text-to-video model called Sora, which is capable of generating realistic and imaginative scenes from a single text prompt. This new technology allows Sora to create lifelike content featuring multiple people, different types of movement, facial expressions, textures, and objects with high detail. Unlike other AI content, the videos generated by Sora do not have a plastic look or nightmarish forms.Sora is also multimodular, meaning users can upload a still image or a pre-existing video and have it animated or extended with attention to small details. OpenAI has shared sample clips on its website and on social media platforms, showcasing the lifelike quality of the generated content. However, Sora is not perfect, as it still has weaknesses. It can have difficulties simulating physics, confuse left from right, and misunderstand cause and effect. The AI also makes funny errors, such as transforming a large piece of paper into a chair and misspelling certain words.
To mitigate potential harms or risks, OpenAI is working with industry experts to assess critical areas and ensure that Sora does not generate false information, hateful content, or exhibit bias. They are also implementing a text classifier to reject prompts that violate their policy, including sexually explicit or violent content. The official launch date for Sora has not been announced yet.
In the meantime, OpenAI's competitors, such as Stability AI and Google, have already released their own video engines. For those interested in AI video editing, TechRadar has compiled a list of the best AI video editors for 2024.