IA: Odyssey shows what the video after-game might look like
The American startup Odyssey presents her interactive experience of the same name. By using a very specific generative AI type type, Odyssey offers Internet users to walk in virtual worlds generated in real time … Without playing engine and without a pre-line environment.
An interactive experience under the background of artificial intelligence
Free access to the Internet for a few days, Odyssey immerses you in an immersive videoin which you are able to move freely by means of the keys of your keyboard. Here, the sets are more like real shots, far from 3D environments modeled typical of classic video games. The experience is intriguing and, although her appearance really prototype, she suggests a promising future for the world of video games.
By means of the interface, you can regenerate new worlds in which you walk virtually. Indeed, unlike a fixed video generated in advance, Odyssey uses a World Model, allowing it to adapt in real time to the actions of the user.
When AI generates real -time environments
In order to understand how the interactive Odyssey experience works, it is important to distribute two types of models used in the generative AI: video models and World Models.
A “classic” video model, which is found on certain platforms such as Sora Or Kling aiworks on the basis of a prompt input, delivering a fixed video output video. Once this one is generated, it is no longer possible to modify it, except by entering a second prompt for this purpose.
A World Model does not generate a video in advance: it uses prediction systems to create each image as and when, depending on the virtual action carried out by the user. For example, when the user presses the key that allows him to move forward, the model generates a new image. It is for this reason that the current experience on the Odyssey site has a rendering in a rather coarse quality.
Operating without a game engine and without pre-generated 3D environment mainly uses data, but above all a phenomenal quantity of GPU.
If the current rendering is limited in terms of interactions, such as the fact that it is not possible to act on any object present in the environment, it nevertheless opens doors in the future of the video game industry.
