You could say Elon Musk’s xAI is ratcheting up the pressure in the AI video world. The company has it that Grok Imagine Video 1.5 will have you making clips at almost twice the speed, with better physics and no more out-of-sync sound. It’s a head-on challenge to the likes of OpenAI and Google as studios are in a hurry to make prompts look like the real thing.
Speed as a strategic lever
xAI says its Fast mode is a lot less of a wait: a 6-second 720p piece of video is done in 25 seconds or so, where it was 40 plus before. Put it all together and you’re looking at 5 to 30 second generation times, which is no small thing when you have to churn through a few versions.
They made a point of it in their announcement – the sharper look, the physics, the pace of it all. Speed is what lets you keep your production cycle tight. And it comes at a time when the field is getting crowded and everyone is trying to put out watchable scenes from a text or an image, in volume.
What actually changed
Now, the model will put down sound effects, some ambience, even dialogue in the same pass as the picture. You don’t have to worry about the two being at odds. They’ve also put some work into the speech for better timing and clarity, in an effort to put an end to the kind of hiccups you see in older AI output.
Then there’s the matter of physics. xAI has things moving with a bit more heft and momentum, so you don’t get as many warps. If you want to put in a camera move in plain English, the system will do as you ask. They call it some of the best you’ll find right now for cinematic control.
As for what you get out the other end, it’s 720p at 24fps, with clips running 6 to 15 seconds. There’s a workflow to extend a video if you need to string a story together in quick beats, without the overhead of a slower model.
Workflow tools and availability
So you can get from a one-off to something you use every day, xAI is rolling out some features. A Projects tab to corral your work from the side, parallel agents to run a few prompts at once rather than put them in a line, and a search to find an old file without the scrolling.
You can get at Grok Imagine Video 1.5 (or ‘grok-imagine-video-1.5’ in the API) and start with an image, set the motion, and pick your resolution and how long you want it. The Fast one is up on grok.com/imagine and in the apps for iOS and Android.
Why this might be the one to put in your cart:
– You can put together a content calendar with faster drafts
– When the audio and video are in step, you have less to fix in post
– Parallel agents mean you aren’t sitting around between prompts
– Search and Projects put your assets in one place
Position versus OpenAI and Google
By xAI’s own numbers, Grok Imagine Video 1.5 is at the top of the Image-to-Video Arena. That’s a 52-point Elo jump on the last version, and it puts them ahead of ByteDance’s Seedance 2.0, HappyHorse 1.0 from Alibaba, and even Google Veo.
The stats back it up: 1.245 billion videos were made in January 2026 alone, and by the first of March 2026, the feature had been visited 314 million times over. It’s powered by the Aurora autoregressive engine and was put through its paces on 110,000 NVIDIA GB200 GPUs.
Still, nailing true realism is an industry-wide problem. But with more stable motion and audio that doesn’t lag, you’re setting a higher standard. It’s only a matter of time before OpenAI and Google have to answer to it, especially when commercial outfits are weighing in on cost and how much rework is involved.
What to watch next
We’ll have to see from the users if the improvements in physics and voice hold up when you get past the demo reel. If they do, then having nearly double the speed and a good sync may well make Grok the go-to for those 6 to 15 second pieces where you can’t afford to dilly-dally.
