Grok Imagine is a set of xAI models that generate images and video in the Grok app and through the API. Video comes from grok-imagine-video-1.5: clips up to 15 seconds, resolution up to 1080p, with sound and speech. Images come from grok-imagine-image-2.0. Through codexkey you can call the video model, while for images the catalog has Nano Banana and GPT Image.
What Grok Imagine is
xAI's documentation describes the Imagine API as a tool for creating and editing images and video. Images come from text and from references, video from text or a photo, and finished clips can be edited and extended (Imagine overview). In the Grok app this lives in an Imagine section, and the Usage tab shows its consumption on a separate line.
Searches for "grok video" and "grok image," in English or in Russian, are mostly about the same thing: drawing or animating something inside Grok. The app and the API differ in limits and prices, covered below. Grok as a chat is the subject of Grok in Russia and Grok API: models and prices.
Grok video: Video 1.5 and 1.5 Lite
As of 29 October 2026 the xAI API has three video slugs: the two main models from the table, plus the base grok-imagine-video for editing and extending. The table is built from xAI's documentation.
| Parameter | grok-imagine-video-1.5 | grok-imagine-video-1.5-lite |
|---|---|---|
| Best for | top quality, references, voices, pinned frames | simple clips at the lowest price |
| Length | 1-15 seconds, 8 by default | up to 15 seconds |
| Resolution | 480p, 720p, 1080p, native 1080p for text and photo | 1080p is an upscale from 720p |
| References | up to 14 images and up to 3 voices | none |
| Sound | native, speech with lip sync | native, speech with lip sync |
| Price per xAI list | from $0.08 per second | from $0.02 per second |
There are five modes: text to video, image to video, references to video, video editing and extension. Editing and extension run through the third, base model grok-imagine-video. The source clip for an edit can be up to 8.7 seconds, an extension adds 2 to 10 seconds, and the result is capped at 720p (video documentation).
Native 1080p for text and photo arrived on 31 July 2026. That day xAI expanded grok-imagine-video-1.5 and added references to video with ready-made voices (release notes).
The price list itself shows how to count the money: the price is set per second and depends on duration and resolution. The minimum for grok-imagine-video-1.5 is $0.08 per second, so from $0.64 for a standard 8-second clip and from $1.20 for a full 15. On Lite the same clips start at $0.16 and $0.30. We derived these sums ourselves from the "from $0.08" and "from $0.02" lines, and the real price depends on resolution. xAI recommends Lite for simple clips and 1.5 where you need references, voices, pinned frames or native 1080p.
The request is asynchronous: you submit a job, poll its status with the returned identifier, then download the clip. The link to the result is temporary, so save the file right away.
The reference mode deserves its own description. In grok-imagine-video-1.5, images define the characters, objects and setting, up to three characters get ready-made voices, and pinned frames (the first, the last and up to four in between) fix the key moments of the scene. In the example from the documentation, two characters in a cafe discuss moving house: there are spoken lines and a change of shot, and the 15-second 720p clip is assembled from references and voices. For now xAI offers your own audio files as a voice only to trusted partners.
For something to compare against, here is Grok Imagine next to other models, per the vendors' documentation:
| Model | Length per pass | References | Voice |
|---|---|---|---|
| Grok Imagine Video 1.5 | up to 15 s | up to 14 images | speech with lip sync, up to 3 voices |
| Veo 3.1 | up to 8 s | up to 3 images | native sound |
| Seedance 2.5 | up to 30 s | up to 50 files | sound together with video |
| Kling 3.0 | up to 15 s | video and images in Omni | speech in several languages |
We covered each of them separately: the AI video generator overview, plus the articles on Veo 3.1 and Seedance 2.5.
Grok image: what grok-imagine-image-2.0 does
Images come from grok-imagine-image-2.0: 1K and 2K resolutions, up to 10 images per request, text-driven editing with references, up to 5 source images. By xAI's price list for 29 October 2026 the price is fixed at $0.04 per image and does not depend on prompt length. An edit is billed twice: for the input image and for the result. xAI retires the old grok-imagine-image-quality slug on 2 November 2026, and requests to it will go to 2.0 at low quality.
Grok Imagine images are not in the codexkey catalog. If images are what you need, take nano-banana, nano-banana-2, nano-banana-pro, gpt-image-2 and gpt-image-2-5 through POST /v1/images/generations. Nano Banana Pro is described in more detail in a separate review.
Is Grok Imagine free: what is known about limits
We found no official figures for free Imagine. This is what xAI's FAQ says:
- Paid plans have one shared weekly limit across all Grok products, Imagine included. Consumption is visible in Settings, on the Usage tab, as percentages per product.
- When the weekly limit runs out, paid features stop until it resets. The FAQ names free limits only for Chat and Voice, and says nothing about Imagine.
- Extra Usage Credits are sold from $5 and for now only on the website (xAI FAQ).
The upshot is simple: you cannot count on free Imagine clips as a standing limit, so check your own account. Access to the app from Russia is a separate question.
How to write a prompt for Grok Imagine
xAI's documentation spells out the mechanics that shape a prompt's style. First, text to video in grok-imagine-video-1.5 works as a chain: the model draws the first frame from the text and only then animates it. So give the start of the prompt to a concrete description of the opening frame, then describe the action. Second, references define characters, objects and setting without pinning the first frame, and a voice can be tied to a character from a ready-made set.
xAI also has a separate guide to image prompts (the Image Prompting Guide in the documentation). For video, three examples of our own will help (originally written in Russian):
- Text to video: "Close-up: a red fox stands in the snow by a fir tree, breath steaming from its nostrils. The fox turns its head to the camera and jumps into a snowdrift. The camera follows it. Sound: crunching snow, wind."
- Dialogue with voices: "Two women in a kitchen, Anna at the stove, Maria at the table. Anna says: 'You're late again,' Maria answers: 'Traffic.' Medium shot, the camera switches over the shoulder."
- Photo to video: "Bring the photo to life: a cup of coffee on a windowsill, rain outside, steam rising over the coffee, the camera slowly moving in."
xAI's documentation does not discuss Russian-language prompts separately, so on your first clip compare a Russian and an English version of the text.
Grok Imagine through the codexkey API
In codexkey the model is called grok-imagine-video, and the request goes to the shared video route: POST /v1/videos, status through GET /v1/videos/{id}, the file through GET /v1/videos/{id}/content. Do not carry addresses over from xAI's documentation. There a video is created with POST https://api.x.ai/v1/videos/generations, while our format is compatible with the OpenAI Videos API (reference).
curl https://api.codexkey.online/v1/videos \
-H "Authorization: Bearer sk_cdx_YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "grok-imagine-video", "prompt": "A red fox on the snow turns its head to the camera and jumps into a snowdrift. Sound: crunching snow and wind"}'
The response carries an id. You put it into GET /v1/videos/{id}, and once the job finishes you download the file. The sk_cdx_... key is created in your account.
Check versions and parameters on the model's page in the catalog. As of 29 October 2026 it lists versions 1.0 and 1.5, durations of 6 and 10 seconds and resolutions of 480p, 720p and 1080p, even though xAI allows 1 to 15 seconds. The catalog can lag behind the vendor's documentation, so check the model page before you send a request. There is no public price list for video: the price is shown on the model page, and money is charged only for a successful result. You can pay in rubles by Russian card, through SBP (the Russian Faster Payments System) or in crypto, with no VPN and no xAI account.
Choosing between Grok Imagine and other video models? Run the same prompt on several and compare the results.
FAQ
Is Grok Imagine free: how much can I generate?
We found no official figures for free Imagine. xAI's FAQ names free limits only for Chat and Voice, while paid plans have one shared weekly limit. Check your own consumption in Settings, on the Usage tab.
How do I make a video in Grok?
In the Grok app, open Imagine, describe a scene or upload a photo. Through the API you send a job to the video model, poll its status and download the clip. In codexkey that is POST /v1/videos with the grok-imagine-video model.
How many seconds of video does Grok Imagine make?
According to xAI's documentation, grok-imagine-video-1.5 makes clips from 1 to 15 seconds, 8 by default. The lighter 1.5-lite also goes up to 15 seconds. The codexkey catalog may narrow the allowed values, so check the model page.
Can Grok make images?
Yes, grok-imagine-image-2.0 handles that: 1K and 2K resolutions, up to 10 images per request, text-driven editing. Those images are not in the codexkey catalog; Nano Banana and GPT Image are available instead.
Does Grok Imagine work in Russia?
The Grok app, its availability and payment are covered in Grok in Russia. The Grok Imagine video model works through a codexkey key with no VPN and no xAI account.
Which Grok Imagine models are current as of 29 October 2026?
Per xAI's documentation: video grok-imagine-video-1.5, grok-imagine-video-1.5-lite and the base grok-imagine-video for editing and extending, images grok-imagine-image-2.0. The grok-imagine-image-quality slug is retired on 2 November 2026.
How does 1.5 differ from 1.5-lite?
Model 1.5 gives native 1080p, up to 14 reference images, voices and pinned frames. Lite is cheaper and suits simple clips, getting its 1080p by upscaling from 720p.