OpenAI, backed by Microsoft, revealed on Thursday its latest project, Sora, a software designed to generate minute-long videos based on text prompts. According to the company, Sora is currently accessible for red teaming purposes, aiding in the identification of flaws within the AI system, and is also available for use by visual artists, designers, and filmmakers to solicit feedback on the model.
In a statement, the company highlighted Sora’s capability to produce intricate scenes featuring multiple characters, specific types of motion, and precise details of both subjects and backgrounds. Moreover, it noted that Sora has the ability to generate multiple shots within a single video.
In addition to video generation from text prompts, Sora can animate still images, as detailed in a blog post by the company.
This latest development follows OpenAI’s previous creation, the ChatGPT chatbot, which gained attention upon its late 2022 release for its proficiency in composing emails, writing code, and crafting poems.
Meta Platforms, the parent company of social media giant Facebook, enhanced its image generation model, Emu, last year by integrating two AI-based features capable of editing and generating videos from text prompts. Meta Platforms is also positioning itself to rival tech giants such as Microsoft, Google (Alphabet), and Amazon in the rapidly evolving landscape of generative AI technology.
While Sora is presented as a work-in-progress, OpenAI acknowledged potential challenges, stating that the model may struggle with spatial details and following specific camera trajectories. Furthermore, the company disclosed ongoing efforts to develop tools capable of identifying videos generated by Sora.