👉 OpenAI's Sora video generator launches today

OpenAI's Sora has officially launched for select users, as highlighted in an early review by YouTuber Marques Brownlee. Available at Sora.com, the platform is designed as a standalone product separate from ChatGPT, showcasing a homepage with recently generated videos. Sora can create videos from both uploaded images and text prompts, with features allowing users to edit existing videos and use a "Re-mix" function to describe desired changes, influencing the final output's artistic style. The model supports video resolutions up to 1080p, though higher resolutions significantly increase generation time, with 1080p taking approximately eight times longer than the fastest option at 480p.

While Sora demonstrates impressive capabilities, it also shares common flaws with other generative tools, such as issues with object permanence and anatomical inaccuracies, particularly concerning legs in walking animations. The platform includes safeguards against generating harmful content or infringing copyrights, and it watermarks videos, although the watermark can be easily cropped out. Overall, Sora is particularly suited for creating animations and abstract content, but it struggles with photorealistic outputs.

👉 World labs creates interactive 3D scenes from photos

World Labs has developed an AI system capable of generating interactive 3D scenes from single images, enabling users to explore environments in a video game-like manner. This technology is part of a new category referred to as "world models," ensuring that generated scenes maintain consistency and adhere to basic physics principles, which enhances realism. While the generated scenes are visually appealing, they currently have limitations, such as restricted movement areas and occasional rendering errors, indicating room for further development.

👉 Learn more here

👉 Google's new AI creates virtual worlds effortlessly

Google's Genie 2 AI model can dynamically generate playable 3D environments in 720p from text prompts or images, simulating world dynamics for a consistent experience. It is designed for training and evaluating embodied agents, showcasing its potential in gaming and simulations. Users can interact with these environments using keyboard and mouse inputs. Project lead Jack Parker-Holder shared impressive videos demonstrating the AI's functionality.

👉 Spotify's wrapped gets an AI twist

Spotify Wrapped 2024 introduces a new AI podcast feature powered by Google’s NotebookLM, which generates personalized audio content tailored to users' listening habits throughout the year. This feature allows users to experience a unique, automated narrative about their favorite songs and genres. The AI creates engaging conversational content that reflects individual listening trends, alongside existing features such as the AI DJ.