OpenAI’s Sora AI Model: A Look at the Latest Generated Videos and What’s Holding Up Its Release
OpenAI has recently unveiled a new set of videos generated using its Sora AI model, showcasing some truly imaginative and fantastical creations. From a horse on rollerskates to a bubble dragon and mythical tea, these videos have been shared on TikTok, sparking excitement and curiosity among viewers.
The Sora AI model has been highly anticipated since its initial reveal in February, with many speculating about when it would be available for public use. However, in a recent interview with Marquees Brownlee’s WVFRM podcast, the Sora team revealed that a public release is unlikely to happen anytime soon. This is due to the need for further safety research and the fact that it takes minutes, not seconds, to generate a video using Sora.
Despite the delay in a public release, the team continues to create captivating videos in response to prompt suggestions from social media users. One of the new videos features a “cute rabbit family eating dinner in their burrow,” showcasing the model’s ability to bring imaginative prompts to life.
What sets Sora apart from other AI video models on the market is its unique architecture, which combines techniques from models like GPT-4 and DALL-E. The team has also utilized a diverse training dataset and a large number of GPUs to train Sora, resulting in stunning and realistic visuals.
While Sora currently relies on text prompts for generating videos, the team is working on adding more fine detail controls to manipulate lighting, camera motion, and orientation. This will give users more control over the output and allow for even more creative possibilities.
Despite the excitement surrounding Sora, its release date remains uncertain. The team is focused on further refining the model and gathering feedback from safety researchers before making it available to the public. So, while we may have to wait a bit longer to play with Sora ourselves, the possibilities it presents for video production, marketing, and architecture are certainly worth the anticipation.
