OpenAI's Sora is launching today — here's highlights from the first review

Sora, OpenAI's new video generator, has launched for select users, as revealed by YouTuber Marques Brownlee. Sora is a standalone product, separate from OpenAI's ChatGPT, and is accessible via Sora.com. It allows users to generate videos from images and text prompts, and offers features like "Re-mix" for editing existing videos and "Storyboard" for creating sequences. The tool can produce up to 1080p resolution videos, though higher resolutions take longer to generate. Brownlee noted that while Sora is useful for creating animations and abstract videos, it struggles with object permanence and anatomical accuracy, particularly with legs.

Sora includes safeguards to prevent the generation of inappropriate content and watermarks each video. It prohibits footage of minors, violence, explicit themes, and recognizable public figures or logos. Despite its innovative capabilities, Brownlee cautioned against using Sora for photorealistic videos, as the AI-generated nature of the clips is often apparent due to inconsistencies and visual anomalies.

Key takeaways:

```html

Sora, OpenAI's video generator, is a standalone product that allows users to create videos from text prompts and uploaded images, with features like "Re-mix" and "Storyboard" for customization and sequencing.
The tool can generate videos up to 1080p resolution, but higher resolutions take significantly longer to process, with 1080p taking 8 times longer than 480p.
Sora faces challenges common to AI video generators, such as issues with object permanence and anatomical inaccuracies, particularly with moving legs.
Sora includes safeguards against generating inappropriate content and watermarks videos, but it is best suited for non-photorealistic content like animations and abstract footage.

```

OpenAI's Sora is launching today — here's highlights from the first review | TechCrunch

Key takeaways:

Comments (0)

Newsletter