xAI's new model animates one photo into video — 'preview' is the word doing the work

On 3 June 2026, xAI dropped grok-imagine-video-1.5-preview via its API: feed it a single still, get back cinematic video with camera moves, atmosphere and physics, 'faithful to the source.' If it delivers what the launch page shows, it's a real entry in the hottest, most-crowded corner of generative AI. The capability story is legitimately interesting. That doesn't mean the words in the headline are doing the same work as the words in the fine print.

Let's be precise about what xAI is claiming and what it isn't. 'Preview' is an early, evolving release where behaviour and quality can change. The announcement is an invitation to try the model, not a certificate of finished product. Generative video models in preview routinely have issues that disappear before GA — temporal flicker, subject drift, inconsistent physics — and it's genuinely unknown how many of those this one carries. That's not a dunk; that's the definition of preview.

'Up to 720p' is the honest ceiling of the current output, and it sits below where the most-established image-to-video systems already operate. Several mature rivals — Runway, OpenAI's Sora, Kuaishou's Kling among them — are generally available and, by their own published claims, reach higher output resolutions than a 720p preview. That gap matters if you're building a product that needs sharp output, and it's worth seeing the shape of the field rather than the launch page in isolation:

Grok Imagine figures are xAI's own; rivals' figures are each company's own published claims as of June 2026. Independent, like-for-like quality comparisons don't yet exist.

Neither of those caveats is a knock on xAI. They're just the accurate read of what 'preview at up to 720p' means, which is useful to know if you're deciding whether to build on this today or wait.

The quieter story is pricing and positioning. xAI says its flagship Grok 4.3 runs at $1.25/M input, $2.50/M output — aggressive numbers relative to many tier-1 peers — and it's now bolting a creative-media line on top. That's a company trying to undercut on text and land in image/video at once. Ambitious strategy; early execution. The architecture of the bet makes sense: a developer who builds on xAI's text API is the obvious next sale for the image and video API. The question is whether a 720p preview is the thing that locks in that relationship or whether developers wait for the GA version with sharper output. Watch what xAI announces as version 1.5 comes out of preview — that is when the actual competitive comparison becomes possible.

An image-to-video model entering at 720p in preview is a credible opening move in a space that still has no settled leader. It's not a category win; it's a round-one punch from a company with the resources to stay in the fight for as long as it takes. The right question for developers evaluating this right now is: does 720p preview quality serve your use case today, or does your product need 1080p GA stability? If the answer is yes to the first, try it; if no, bookmark the GA announcement. Either way, xAI's creative-stack trajectory is worth following — this is clearly not the last move they make here.