Grok Imagine Video 1.5 cuts 720p generation from 40+ seconds to 25 seconds

xAI is pushing Grok Imagine further into the AI video race with a faster image-to-video model and general API availability. In its June 17 launch tweet, xAI wrote: “Grok Imagine Video 1.5 is here. Our new image-to-video model with sharper realism, better physics and faster generations.” The source tweet is here.

The most useful product detail came in the same thread: “720p videos now render in about 25 seconds, down from 40+ in our previous model.” xAI’s linked news post says Grok Imagine Video 1.5 Fast produces 6-second, 720p clips in about 25 seconds, nearly doubling generation speed. The post also says Video 1.5 Fast is available on grok.com/imagine and xAI’s iOS and Android apps, while Imagine Video 1.5 is generally available in the xAI API as grok-imagine-video-1.5.

The release is framed around three areas: audio and speech, motion and physics, and workflow. xAI says sound effects, ambience, and dialogue are generated in the same pass and are better synced to the action. It also claims movement holds together over the length of a clip with fewer warps and more believable weight and momentum. For developers, the API takes a starting image, a motion description, resolution, and duration.

xAI’s official account usually posts Grok model releases, API updates, and platform integrations. This tweet is material because it moves Imagine Video 1.5 from preview-style positioning into broader consumer and API access, with a concrete speed comparison. The 25-second figure is especially important for creators and app developers who iterate through many generations before picking one result.

The next thing to watch is whether the quality claims hold outside xAI’s examples. AI video models often look strongest in curated demos, while real user prompts expose temporal drift, object warping, inconsistent identities, and weak audio alignment. For developers, pricing, rate limits, moderation behavior, and latency under load will matter as much as the model name.