The company xAI has made the Grok Imagine Video 1.5 model publicly available. The tool is an image-to-video generator. The developers placed the main emphasis on realism, correct motion physics and high rendering speed.

The model has officially left preview mode in the API. At the same time, users gained access to an accelerated version of the technology.

What the new model can do

The Grok Imagine Video 1.5 service is available in the Imagine API under the identifier grok-imagine-video-1.5. In parallel, the team rolled out the Video 1.5 Fast version. It works in the grok.com/imagine web service, as well as in the iOS and Android apps.

The main speed gain applies specifically to the Fast version. xAI claims an almost twofold acceleration of work.

The quality improvements affected several important areas at once. Audio accompaniment, background atmosphere and lines are now created in a single pass. All the elements are perfectly synchronized with the action in the frame, and the characters' speech has become more expressive and clearer.

The movement of objects becomes more stable throughout the entire clip. Fewer deformations are noticeable in the frames, and weight and inertia look more believable. Through the API, developers can set a starting image, a text description of the movement, resolution and duration.

Along with the model, xAI is consistently introducing useful tools for the workflow. Projects in the side panel, parallel launching of several generations and convenient search through one's own library of materials have appeared in the interface.

How users reacted

Part of the audience tested the model for several days. The first users rated the result positively, calling the team's work high quality. But there is a fly in the ointment, too. For example, in the promotional material attentive viewers noticed an unfortunate error in the product's name.

Many commenters formulated their wishes. Among the main requests, adding voice references stands out. The function is needed for creating characters with a consistent sound, as in competing services.

Source: BeInCrypto