The short answer
- Grok Imagine video is xAI's video model line for text-to-video, image-to-video, reference-to-video with up to 14 images, lip-synced speech, editing and extension, in clips up to 15 seconds.
- API prices: 1.5 at $0.08 (480p), $0.14 (720p) and $0.25 (1080p) a second; Lite from $0.02 a second.
- Vertical 9:16 is supported, and image-to-video matches the input image's aspect ratio.
- Reference-to-video, editing and extension top out at 720p. Check on-screen text before running any clip.
Prices and limits as listed on xAI's pricing page and developer docs on November 9, 2026.
The Grok Imagine video models
Grok Imagine is xAI's image and video model line. Its video side has three models in the developer docs, and which one you use decides what you can do. All figures here are from xAI's developer docs and pricing page, checked on November 9, 2026.
| Model | What it does | Resolution | Price per second |
|---|---|---|---|
| grok-imagine-video-1.5 | Text-to-video and image-to-video with lip-synced speech; up to 14 reference images and 3 voice references; pinned first, last and mid-video frames | 480p, 720p, native 1080p | $0.08 (480p), $0.14 (720p), $0.25 (1080p) |
| grok-imagine-video-1.5-lite | Text-to-video and image-to-video with lip-synced speech, the lowest price per second; 1080p is upscaled from 720p | 480p, 720p, 1080p | $0.02 (480p), $0.03 (720p), $0.14 (1080p) |
| grok-imagine-video | Video editing and video extension; reference images (up to 7) | 480p, 720p | $0.05 (480p), $0.07 (720p) |
Clips run from 1 to 15 seconds on all three. Aspect ratios include 9:16, 1:1, 16:9, 4:3, 3:4, 3:2, 2:3, 21:9 and 5:2, which covers every paid-social placement. Image-to-video matches the aspect ratio of the input image automatically. Image inputs carry a small extra fee of $0.01 per image on the 1.5 models, and requests are asynchronous: you submit, poll and download.
xAI's own guidance is simple. Use the Lite model for simple videos, and the full 1.5 model when you need references, voices, pinned frames or native 1080p.
What each mode does for an ad
Image to video: animate the packshot
Give it a clean product still and a motion instruction, such as "slow push-in, steam rises from the cup, soft window light". This is the safest route for product ads, because the first frame is your real product. Pair it with an image model; the prompt formats in Nano Banana prompts for product ads make good start frames, and image to video AI covers the method in general.
Reference to video: keep the product and the person consistent
The 1.5 model takes up to 14 reference images and says they supply characters, products and locations without locking the first frame. xAI names virtual try-on, product placement and recurring characters as the use cases. You point to each image by position in the prompt (the first is IMAGE_0, then IMAGE_1). xAI's advice for many references is to give each one a single role and a fixed place in the scene, keep the cast small and keep props static. Reference-to-video is capped at 720p.
Speech and voices
The 1.5 models generate lip-synced speech. On the full model, up to three characters can be given preset voices. For ads this means a talking spokesperson clip in one generation, without a separate voice track. Always play the audio and read the captions against your script; a lip-synced line can still say the wrong word.
Pinned frames
You can set the first frame, the last frame and up to four mid-video keyframes. In ad terms that is a storyboard: frame one is the problem, the last frame is the product with the offer. It is the nearest thing in the lineup to directing a shot rather than hoping for one.
Editing and extending an existing clip
The older grok-imagine-video model edits a source clip up to 8.7 seconds with a text instruction, such as changing the colour of an outfit or adding an object, and keeps the rest close to the source. It can also extend a clip by 2 to 10 seconds. Output is capped at 720p. xAI's January 28, 2026 announcement reports human raters preferred Grok Imagine's edits over Kling O1 and Runway Aleph on its own benchmark, which is xAI's evaluation and not an independent one.
What a Grok Imagine ad costs
At the 720p price of $0.14 a second, a 10 second clip is $1.40. At 1080p it is $2.50. On Lite at 720p it is $0.30. xAI's own Product Photo to Video Ad cookbook builds an eight second vertical ad from one product photo, using three generated scenes and one video, and reports about $1.35 for the run. Plan for several runs per concept, because the first take is rarely the one you keep.
| Clip | Model and resolution | Cost |
|---|---|---|
| 6 second hook | Lite, 720p | $0.18 |
| 10 second spokesperson | 1.5, 720p | $1.40 |
| 15 second hero ad | 1.5, 1080p | $3.75 |
| 15 second hero ad, three takes | 1.5, 1080p | $11.25 |
How it compares
xAI's launch post cited Artificial Analysis and LMArena text-to-video rankings in which Grok Imagine led on January 28, 2026, and leaderboards move monthly. For ad work, compare on briefs rather than rank: see the best AI video generators for ads for how Seedance, Kling, Veo, Wan and MiniMax behave on faces, product shots and brand text. Grok Imagine 1.5 is also available inside studios; the MakeUGC review covers one that lists it next to the others on one credit balance.
Limits to plan around
- Reference-to-video tops out at 720p, and editing and extension are on the older model at 720p.
- Editing sources are short. 8.7 seconds for edits.
- Brand text and labels still need checking. Treat any on-screen word as unverified until you read it.
- It is an API and a model, not a campaign tool. You get clips. Targeting, testing and spend are separate.
Where Sprites fits
Choosing a model, writing its prompts and stitching clips is work most teams would rather not own. Director, the video ads add-on from Sprites, takes a brief and returns finished video ads, and Sprites runs the campaigns that carry them with every change held for your approval. See the pricing page; plans start with a 7-day trial for $1.
Frequently asked questions
What is Grok Imagine video?
Grok Imagine video is xAI's video generation model line. It turns text, a still image, reference images or an existing clip into video with generated audio, in clips up to 15 seconds, and is available through xAI's API.
How much does Grok Imagine video cost?
As listed on xAI's pricing page on November 9, 2026, grok-imagine-video-1.5 costs $0.08 a second at 480p, $0.14 at 720p and $0.25 at 1080p. The Lite model is $0.02, $0.03 and $0.14, and the older grok-imagine-video model is $0.05 at 480p and $0.07 at 720p.
Can Grok Imagine make 9:16 vertical ads?
Yes. The docs list 9:16 among nine supported aspect ratios, along with 1:1 and 4:3 for feed placements. Image-to-video automatically matches the input image's aspect ratio.
Does Grok Imagine generate speech and audio?
Yes. The 1.5 models generate native audio with lip-synced speech, and the full 1.5 model accepts up to three preset voice references. Check the audio against your script before publishing.
Can Grok Imagine keep my product consistent across shots?
The 1.5 model accepts up to 14 reference images that supply products, characters and locations, and can pin first, last and mid-video frames. xAI lists product placement and virtual try-on as use cases. Reference-to-video output is capped at 720p.
Which Grok Imagine model should I use for ads?
xAI recommends Lite for simple videos and the full 1.5 model when you need references, voices, pinned frames or native 1080p. For editing or extending an existing clip, use the older grok-imagine-video model.