xAI's Grok Imagine 1.5 turns a still into video — and stakes a claim in the generative-video race

On 3 June 2026, xAI made grok-imagine-video-1.5-preview — its latest image-to-video model — available in preview via the xAI API. The pitch, per xAI, is straightforward and ambitious: feed it a single still image and it returns fluid, cinematic video, complete with camera moves, atmosphere and a sense of physics, while staying faithful to the source — at up to 720p. It is the latest marker in a rapidly crowding space, and it carries the two-word caveat that matters most: in preview.

Image-to-video is one of the most competitive frontiers in generative AI right now. The hard problem is not producing motion — that is table stakes — but producing motion that remains faithful to the source image: a portrait that stays recognisable, a landscape that moves without distorting. xAI frames Grok Imagine 1.5 as doing exactly that, adding camera drift, atmospheric detail and physical plausibility to a static frame. These are the model's own claimed properties; the 'preview' label is xAI's signal that quality and behaviour are still evolving. Until independent benchmarks exist, the capability claims should be treated as the starting position for a conversation, not the final word.

The June release does not arrive in isolation. In May, xAI shipped a 'Quality Mode' for image generation and editing, available to enterprise and teams via the Grok Imagine API, per the May announcement. Read in sequence, the picture is of a deliberate creative-media stack being assembled rung by rung: first a quality tier for stills, then a preview video model. That is a coherent product arc, even if each individual release is early and incomplete. The API-first approach — no consumer app, developer-targeted — is consistent with how xAI has built across its product lines: ship to builders, let them experiment, iterate from real usage.

The image-to-video category has moved fast in 2025–26. The current landscape includes capable systems from Runway, Kuaishou's Kling, OpenAI's Sora and others — models that compete on resolution, clip length, motion quality and faithfulness to the prompt or source. The cleanest way to place this release is by stage of maturity rather than a spec-by-spec leaderboard, since the only first-party figures we have are xAI's own:

Grok Imagine figures are xAI's own (3 June 2026); rivals' figures are each company's own published claims. No independent, like-for-like benchmark of these systems exists yet.

Grok Imagine 1.5 enters at 720p in preview — respectable for an early release, but below the higher-resolution output the field's more mature systems claim and where they already ship generally. The more relevant question is not today's resolution ceiling but how xAI's rate of improvement compares to the field's in the coming quarters.

For context on xAI's broader positioning — and attributed to xAI — the flagship language model Grok 4.3 is priced at $1.25/M input and $2.50/M output, with configurable reasoning effort. That puts text inference at aggressive pricing relative to many peers. Whether xAI follows the same strategy on creative-media pricing, and whether competitive rates draw developers to Grok Imagine when GA models elsewhere are entrenched, is the commercial question worth tracking.

The immediate question is whether Grok Imagine 1.5 exits preview as a finished, stable product or iterates significantly before reaching general availability. Preview releases in generative media frequently ship with motion artefacts, temporal inconsistency and prompt-drift that improve substantially between preview and GA. The second question is pricing: xAI's text-inference pricing is among the more competitive in the market, and whether that extends to video generation will matter to developer adoption decisions. The third — and most durable — question is whether the 'one API, text and creative-media' pitch is coherent enough to pull developers away from point solutions they already know. None of those questions have answers yet. What is clear is that xAI is moving deliberately, not randomly, toward a full creative toolkit, and the June release is a meaningful step in that direction.