Choose a category
3 models available
4 models available
3 models available
2 models available
4 models available
8 models available
Grok Imagine 1.5 — Image to Video
Reference images that lock a character's identity across shots.
Creative Vision
Describe the motion. Name the reference images to keep a character consistent.
Video Parameters
Priced by the second.
Cinema Preview
Generation History0 records
Grok Imagine 1.5 Image to Video Generator
The 1.5 preview of xAI's Grok Imagine takes up to three reference images instead of one, which is what makes identity lock possible: name a reference in the prompt and the same face carries through the shot. One to fifteen seconds, 480p or 720p, from 1 credit.
Open the prompt with "Reference Image 1 as identity lock" when a person has to stay recognisable
Why Choose Grok Imagine 1.5 Image to Video
This is a different model from the Grok Imagine page you may already have used, not a newer button on the same one. The difference worth crossing over for is the reference image count: 1.5 accepts several, and the original accepts exactly one.
Identity Lock Across the Shot
Upload up to three references and point at them in the prompt — "Reference Image 1 as identity lock" — and the model holds the same person through changes of angle, lighting and setting. One image cannot pin a face down the way three can.
Up to Three Reference Images
JPEG, PNG or WebP, up to 20MB each. They are references, not a first frame — the model reads them for who and what is in the shot rather than starting the video from one of them.
From 1 Credit a Clip
480p is billed per second and starts at a single credit. Eight seconds at 480p is 5 credits, the full fifteen at 720p is 20. Among the cheapest ways on this site to test an idea with a real face in it.
Aspect Ratio That Actually Applies
With more than one reference image you get a real choice of frame — 16:9, 9:16, 1:1, 3:2, 2:3 or auto. The original Grok Imagine image-to-video has no ratio control at all because it only ever takes one image.
One to Fifteen Seconds
A continuous slider from a single second upwards, so a one-second reaction shot is a legitimate and very cheap thing to generate. Priced by the second throughout.
Failed Runs Are Refunded
Credits are held when you submit and returned automatically if the generation fails or the provider never answers. A clip that does not arrive costs you nothing.
Three Images, One Face
The reason to use this page instead of the original Grok Imagine is a character who has to stay recognisable. If you only have one photo and one shot, the original is longer and just as cheap.
How to Use Grok Imagine 1.5 Image to Video
Reference images only work if the prompt refers to them. The steps below are mostly about that link — uploading three photos and then writing a prompt that never mentions them wastes the feature entirely.
Upload One to Three References
For a person, three angles of the same face beat three different people. JPEG, PNG or WebP at up to 20MB each. Upload order matters: the first file is Reference Image 1, the second is 2, the third is 3.
Name Them in the Prompt
Write "Reference Image 1 as identity lock" or "the woman from Reference Image 2 walks into frame". Referring to them by number is how the model knows which upload governs which element of the shot.
Test Short at 480p
Three or four seconds at 480p costs 2 or 3 credits. That is enough to see whether the face survived and whether the motion is the motion you asked for — the two things that go wrong most often.
Extend at 720p
Same prompt, longer slider, higher resolution: 9 credits for eight seconds at 720p, 20 for the full fifteen. Generation is quick on this model — usually under a minute or two — and the clip lands in your history.
Auto Ratio With a Single Image
With exactly one reference the aspect ratio setting has no effect and the output follows your image. Leave it on Auto unless you have uploaded two or three.
Grok Imagine 1.5 Image to Video
FAQ
Reference images, identity lock, clip length, resolution and credit costs for the Grok Imagine 1.5 preview model.
They are two separate models at the provider, not two views of one. This one is the 1.5 preview: it accepts up to three reference images, offers a working aspect-ratio control, and runs 1 to 15 seconds at 480p or 720p. The original accepts a single image, has no ratio control, and runs 6 to 30 seconds with a 1080p tier and a fun-mode style switch.
It is a prompting technique rather than a setting. You upload photos of a person and then tell the model to treat one of them as the identity to hold — "Reference Image 1 as identity lock" — so the same face persists as the shot changes angle or lighting. More reference images give the model more to work from.
Between 1 and 20 credits. At 480p: 1 credit for a single second, 3 for five seconds, 5 for eight, 9 for fifteen. At 720p: 2, 6, 9 and 20 for the same lengths. The cost for your current settings is shown above the generate button before you submit.
Two reasons. The provider publishes a rate only for the image-to-video path on this model, so a text-only run would be priced on a guess. And without images it is simply a shorter, dearer version of the original Grok Imagine text-to-video page, which is right here on the site and better at that job.
The provider documents that multiple files are accepted but publishes no upper limit, so three is a cap we set ourselves rather than one the model imposes. It is deliberately conservative: submitting more than the model actually accepts would fail the run rather than degrade it.
No to both — this model offers 480p and 720p, and the fun/normal style switch belongs to the original Grok Imagine. If you want either of those, use the Grok Imagine image to video page, which has the style switch and runs longer clips.
This endpoint exposes no audio parameter. For video with dialogue and effects, Kling 2.6 and Seedance 1.5 Pro on this site both generate sound alongside the picture.
You are not charged. Credits are held when you submit and released back to your balance if the model fails or the provider never returns a result. Generated videos stay available for seven days, so download anything you want to keep.
Try Grok Imagine 1.5 Image to Video
Three reference images, four seconds at 480p, and 3 credits to see whether the face holds.