In the realm of artificial intelligence, the much-anticipated unveiling of OpenAI’s Sora video generator model has finally come to pass. After months of eager anticipation, the company has at last revealed a working version of Sora as part of its festive “12 Days of OpenAI” event. OpenAI CEO Sam Altman remarked during the live stream that this momentous occasion marks a pivotal step towards achieving the company’s ultimate goal of Artificial General Intelligence (AGI).
Diving deeper into the intricacies of Sora, it is noteworthy to mention that the model will initially be accessible to Plus and Pro subscribers in the United States and worldwide. Notably, renowned YouTuber Marquis Brownlee was granted early access to the video generator, providing a brief review on his channel. Surprisingly, Sora does not rely on the foundation of GPT-4 like its predecessors, standing as a unique addition to OpenAI’s generative tools. While the model is not yet accessible via the standard ChatGPT website, it can be found on Sora.com, which is still awaiting its official launch.
What sets Sora apart from its predecessors is its capability to craft videos ranging from 480p to 1080p in resolution and 5 to 20 seconds in length, all originating from text prompts or reference images. Further fueling the creative potential, Sora can manipulate and lengthen existing video clips. ChatGPT Plus subscribers can generate up to 50 clips each month, with Pro users enjoying unlimited generations and enhanced resolutions and durations. Additionally, Sora introduces a novel “storyboard” feature, enabling users to amalgamate various prompts into a cohesive cinematic scene.
However, as Marquis Brownlee pointed out in his review, Sora’s processing time for a 1080p clip may pose a slight inconvenience, particularly when user activity surges. Notably, the model exhibits challenges in accurately depicting leg movements, resulting in unnatural and perplexing visualizations. Nevertheless, the possibilities unlocked by Sora are vast, promising a realm of creative potential for users worldwide.
In contrast to the unbridled creativity encouraged by its predecessor Grok 2, Sora enforces limitations to safeguard against copyright infringement and the creation of objectionable content. Overcoming a trail of setbacks and delays, OpenAI’s journey with Sora has been fraught with challenges, allowing rival models like Runway’s Gen-3 alpha and Kuaishou Technology’s Kling to take the lead in the AI landscape. Despite these hurdles, the unveiling of Sora beckons a new era of AI innovation, promising boundless opportunities for creators and innovators alike.
