OpenAI, the renowned artificial intelligence research organization, has officially launched Sora, its highly anticipated video generator. The news was revealed by popular YouTuber Marques Brownlee, who got early access to the platform and shared his initial impressions in a 15-minute review.

Sora, accessible at Sora.com, is a standalone product separate from OpenAI's AI-powered chatbot platform, ChatGPT. The website features a scroll of recently generated and curated Sora videos, which can be bookmarked, organized into folders, and viewed with their corresponding text prompts.

One of Sora's most impressive features is its ability to generate videos from uploaded images as well as text prompts. Additionally, users can edit existing videos using the platform. The "Re-mix" feature allows users to describe changes they want to see in a video, and Sora will attempt to incorporate these changes in a newly generated clip. This feature comes with a "strength" setting, enabling users to specify how drastically they want Sora to alter the target video.

In terms of video quality, Sora can generate up to 1080p footage, although higher resolutions come with longer generation times. Brownlee reported that 1080p videos take around 8 times longer to generate than 480p videos, the fastest option, while 720p videos take 4 times longer. The average 1080p video took a "couple of minutes" to generate during Brownlee's testing, but he expressed concerns about how much longer it might take when the platform is open to the general public.

Sora also features a "Storyboard" tool, which enables users to string together prompts to create a scene. This is designed to improve consistency, a common weakness in video generators. However, Brownlee noted that Sora still struggles with objects passing in front of or behind each other in unrealistic ways, as well as legs and other anatomical issues.

OpenAI has implemented several safeguards in Sora, including prohibiting the generation of footage showing people under 18, containing violence or explicit themes, or infringing on copyrighted material. The platform also won't generate videos from images featuring public figures, recognizable characters, or logos, and each video is watermarked.

According to Brownlee, Sora is useful for tasks like creating title slides, animations, abstracts, and stop-motion footage, but it falls short in producing photorealistic content. While the platform's capabilities are impressive, the generated videos often appear "wonky" and lack the realism of human-created content.

The launch of Sora marks a significant milestone in the development of AI-powered video generation technology. As the platform continues to evolve and improve, it will be interesting to see how it is adopted by creators and the potential impact it may have on the video production industry.