Grok Imagine Video 1.5: xAI's video model — sound, 14 references, voices and keyframes

What is Grok Imagine Video 1.5

Grok Imagine Video 1.5 is a video generation model from xAI, the company that makes the Grok chatbot. It launched in two stages:

• June 16, 2026 — image-to-video and the faster Video 1.5 Fast in the Grok apps; the API model left preview;
• July 31, 2026 — reference images and voices, text to video and 1080p.

In October 2026 it was joined by the lighter Video 1.5 Lite. The model runs on grok.com/imagine, in the Grok apps for iOS and Android and through the xAI API as grok-imagine-video-1.5. Images in the same ecosystem come from Imagine Image 2.0 — both models are covered on the Grok Imagine page.

What Grok Imagine Video 1.5 can do

Clips with sound, 1–15 seconds. Sound effects, ambience and lip-synced speech are generated in the same pass as the action. You can turn sound off for a silent clip.

Native 1080p. From text or a photo, the model renders 1080p directly, with no upscaling. The default is 480p, meant for drafts. In reference mode the ceiling is 720p.

Nine aspect ratios. 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3 and the less common 21:9 and 5:2 — for banners and wide covers. When animating a photo, the aspect ratio follows the source image.

Physics and motion. xAI calls 1.5 its best motion model yet: fewer warps and more believable weight and momentum.

Text to video, via a frame. If the prompt has no image, the model first draws a starting frame and then animates it.

One character across scenes. Up to 14 reference images keep the character’s face and look consistent. Voice is set separately: up to 3 speaking characters per clip, each with its own voice from a preset library.

First, last and keyframes. You can pin the exact last frame, and with a source image the first frame too. In between, you can place up to 4 keyframes, each tied to a moment in the clip on a 1/3-second grid.

With images, references or keyframes, the prompt is optional: the model can take the motion from the frames themselves. Describing the action and sound still gives more predictable results.

Video 1.5 vs. Fast vs. Lite

	Video 1.5	Video 1.5 Fast	Video 1.5 Lite
Where	apps and API	Grok apps	API
Speed	a complex clip can take several minutes	6 s at 720p in about 25 seconds	faster than the full model
1080p	native	—	upscaled from 720p
Length	1–15 s	—	1–15 s
References and voices	14 images, 3 voices	—	none
First, last and keyframes	yes	—	no
API price	$0.08 per second	—	$0.02 per second

Fast is a mode in the Grok apps: xAI says it’s nearly twice as fast as the previous model, which needed more than 40 seconds for a 6-second 720p clip. Lite is a separate API model for simple clips from a single prompt or image.

What Grok Imagine Video 1.5 is good for

• Short, high-energy clips: sports, animals, stunts, slow motion at the right moment.
• Series with one character and voice — up to 14 references and 3 voices.
• Clips with a set ending: first and last frames fix how the scene begins and ends.
• Complex paths — a flying object, a camera passing through rooms — from 4 keyframes.
• Animated 21:9 and 5:2 banners and covers.

Grok Imagine Video 1.5 vs. alternatives

	Grok Imagine Video 1.5	Kling 3.0	Veo 3.1	Seedance 2.0
Developer	xAI	Kuaishou	Google DeepMind	ByteDance Seed
Length per generation	1–15 s	up to 15 s	4, 6, 8 s	4–15 s
Max resolution	1080p	4K	4K	4K
References	14 images, 3 voices	up to 7 images or elements	up to 3 images	9 images, 3 videos, 3 audio
Keyframes	first, last and up to 4 in between	first and last	first and last	first and last
21:9 and 5:2	yes	—	—	21:9
In AI2Face	yes	yes	yes	yes

Grok Imagine Video 1.5 stands out with in-between keyframes and its number of image references. For 4K and long dialogue, go with Kling 3.0 and Veo 3.1; for references from video and audio, Seedance 2.0.

How to write prompts for Grok Imagine Video 1.5

1. Describe the action in order: what happens at the start, middle and end of the clip.
2. Add the camera: “the camera follows from behind,” “slow motion at the peak of the jump,” “orbit.”
3. Put sound on its own line: dialogue in quotes, effects, music or no music.
4. With keyframes, describe the transitions between them rather than the frames themselves — the images already define those.
5. Draft at 480p, render the final at 720p or 1080p.

A prompt for the keyframes above:

A boy launches a paper airplane from the back row. It glides smoothly over the desks,
turns toward the sunlit window, catches a draft and lands softly on the teacher's desk
next to a globe. The camera tracks the plane from the side and pushes in on the desk at the end.
Sound: rustling paper, muffled laughter from the class, birds outside, no music.

How to use Grok Imagine Video 1.5 in AI2Face

Grok Imagine Video 1.5 is in the AI2Face generator next to Kling 3.0, Veo 3.1 and Seedance 2.0.

1. Sign up — new accounts get starter credits.
2. Open the generator, switch to video and pick Grok Imagine Video 1.5.
3. Upload a photo, references or keyframes, write the prompt, set the length and resolution.
4. The credit cost is shown before you start. Credits don’t expire.

For a step-by-step look at modes and limits inside Grok itself, see our Grok video guide. Keyframes and character references are easy to create with Nano Banana Pro in the same generator.

Limitations

• 720p with references. Native 1080p only applies to text-to-video and image-to-video.
• No editing or extension in 1.5. xAI handles clip editing with the older Grok Imagine Video model, at 720p and for clips up to 8.7 seconds.
• Preset voices. In the API, each of the 3 speaking characters gets a voice from xAI’s preset library.
• Generation time. A complex high-resolution clip can take several minutes.
• Moderation. Requests and results are filtered under xAI’s policies; the API flags a blocked clip in its response.

Bottom line

Grok Imagine Video 1.5 is a strong choice for short, high-energy clips and scenes where the motion path has to be exact: in-between keyframes are unique to it. Fourteen references and three voices let you carry one character through a series. If you need 4K or long dialogue, compare it with Kling 3.0 and Veo 3.1. For simple clips from a single photo, Video 1.5 Lite is cheaper.