xAI video model · official ID grok-imagine-video-1.5
Grok Imagine Video 1.5 Video Generator
Run the model registered as grok-imagine-video-1.5. This page opens text-to-video and image-to-video; the reference, voice, and last-frame workflows the vendor documents are listed below with their exact boundaries.
grok-imagine-video-1.5-preview and grok-imagine-video-1.5-2026-05-30
Developer
xAI
Input modalities
Text, image, and audio
Output
Video, delivered with an audio track unless silence is requested
Official clip length
1 to 15 seconds, default 8
Official resolutions
480p, 720p, and 1080p
Official aspect ratios
1:1, 16:9, 9:16, 4:3, 3:4, 3:2, and 2:3
Official price per second
0.08 USD at 480p, 0.14 USD at 720p, 0.25 USD at 1080p
Official input prices
0.01 USD per image, text free, preset-voice audio free
Official regions
us-east-1 and us-west-2
Rate limit
10 requests per second
Availability dates
Generally available 16 June 2026; image references and native 1080p added 31 July 2026
Page route
/models/video/grok-imagine-video-1-5/
Official capabilities beside the controls this page exposes
Two boundary lists, deliberately kept apart: what the vendor documents for this model ID, and what this workspace actually opens.
OFFICIAL
Documented by the vendor
Text-to-video from a prompt alone
Image-to-video from a single source image
Reference-to-video using reference images, preset voices, or both, with up to three preset voices per request
First and last frame pinning through the last_frame field
480p, 720p, and 1080p output, with 1080p scoped to text-to-video and image-to-video and reference-to-video capped at 720p
Clip length from 1 to 15 seconds, default 8
Seven aspect ratios, including 3:2 and 2:3
An audio track by default, switchable off with generate_audio=False
Video editing and video extension as separate endpoints, which the vendor documents against the classic grok-imagine-video model rather than this one
LIVE PLATFORM
Open on this page
Text-to-video
Image-to-video with one uploaded image
Quality limited to 720p and 1080p
Clip length from 2 to 15 seconds, default 5
Five aspect ratios: 16:9, 9:16, 1:1, 4:3, and 3:4
A point estimate shown before submission, plus creation history for the current session
Two of these boundaries cross rather than nest, and they are the ones most often misread. Clip length: the vendor accepts from 1 second and defaults to 8, while this page accepts from 2 seconds and defaults to 5. Resolution: the vendor lists 480p as its default and this page has no 480p option at all. Everything else in the first list sits above the second — reference images, preset voices, last-frame pinning, 3:2 and 2:3 ratios, and the audio switch are documented capabilities of the model ID that are not exposed as controls here, so nothing on this page should be read as offering them.
Model brief
What Grok Imagine Video 1.5 changes
Grok Imagine Video 1.5 is the video model xAI registers under the model ID grok-imagine-video-1.5, with aliases grok-imagine-video-1.5-preview and grok-imagine-video-1.5-2026-05-30. It left preview and became generally available on the Imagine API on 16 June 2026. The vendor describes the model as taking text, image, and audio as inputs and returning video, and the release notes register it as the successor to the earlier grok-imagine-video.
The 31 July 2026 update changed what the model accepts. Image references, text-to-video, and native 1080p all went live in the API that day, and voice references were opened on request. The effect is that the same model ID now serves four distinct request shapes rather than the single image-to-video route it launched with, and text-to-video is not a separate pipeline: the vendor documents it as running text-to-image and then image-to-video behind one request, with the intermediate still never returned.
The parameters are worth reading together rather than one at a time, because the boundaries interact. Clip length is an integer from 1 to 15 seconds with a default of 8. Resolution accepts 480p, 720p, or 1080p, and the documentation marks 480p as the default. Native 1080p is scoped to text-to-video and image-to-video only; reference-to-video is capped at 720p. Seven aspect ratios are accepted, and image-to-video otherwise inherits the source frame ratio unless the aspect_ratio field overrides it.
This page is an independent route onto that model, and the workspace above does not expose every documented capability. Text-to-video and image-to-video are open here; reference images, preset voices, and last-frame pinning are documented by the vendor and are described below as vendor-side capabilities rather than as controls you can operate on this page. Where the two boundaries differ, the section headed with the two lists sets them side by side.
03 / Capabilities
Documented inputs the Grok Imagine Video 1.5 API accepts
Each item below is taken from the vendor model card, the Videos API reference, or the video capability and release-note pages, all checked on 13 September 2026.
01
Text-to-video from a prompt alone
A prompt with no image is enough. The vendor documents the mechanism as text-to-image followed by image-to-video, driven by a single request; the intermediate frame is generated internally and is not returned to you.
02
Image-to-video from one source frame
Supply an image and, optionally, a prompt. The vendor documents the prompt as optional on this route — send the image on its own and the model animates it without further instruction. The output inherits the source ratio unless aspect_ratio is set explicitly.
03
Reference-to-video with images and preset voices
Reference images, reference audio, or both select this mode. Up to three preset voices can be attached per request, and the vendor announcement describes image references as each locking one element in place, so a character can be held while the scene changes. This route is capped at 720p.
04
First and last frame pinning
On this model ID, last_frame pins the closing frame and the clip ends by arriving on that image rather than re-rendering it. Pairing image with last_frame pins both ends and lets the model interpolate between them. Classic grok-imagine-video rejects last_frame outright.
05
Audio generated in the same pass
The vendor describes sound effects, ambience, and dialogue as produced together with the picture and landing on the action, with speech clearer and better synchronised than the previous generation. An audio track is on by default; generate_audio=False requests a silent clip instead.
06
Seven aspect ratios and three resolutions
Accepted ratios are 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, and 2:3, defaulting to 16:9. Resolutions are 480p, 720p, and 1080p, with 480p as the documented default and 1080p limited to the two frame-based workflows.
07
Asynchronous requests with a request ID
A generation call returns a request_id rather than a finished file. You poll the request until its status reads done and a video URL comes back, or until it reports expired. The vendor lists prompt complexity, clip length, resolution, and whether editing is involved as the factors that decide how long that takes.
08
Persisted output and a published price list
Generated files are handed back as temporary URLs, and storage_options with expires_after can retain a copy for up to 2,592,000 seconds, which is 30 days. The model card publishes per-second prices by resolution, so a clip budget can be read directly off the official page.
04 / Scene ideas
Jobs Grok Imagine Video 1.5 fits
Each item below is taken from the vendor model card, the Videos API reference, or the video capability and release-note pages, all checked on 13 September 2026.
00:00:00SHOT 01Storyboard placeholder
Product motion from a single still
One clean product photograph becomes a short moving shot: an orbit, a slow push, or a reveal. Keep the label and the casing proportions fixed, and reject any take that redraws the artwork.
00:00:02SHOT 02Storyboard placeholder
Character shorts with a consistent face
The vendor documents reference images as each locking one element in place, so a character can be carried across scenes with the scene swapped around them. That workflow is documented for this model ID and is not open as a control on this page.
00:00:04SHOT 03Storyboard placeholder
Dialogue and voice-driven pieces
Audio is generated in the same pass as the picture, and up to three preset voices can be attached through reference-to-video. Voice reference beyond the preset roster is described by the vendor as available on request rather than generally available.
00:00:06SHOT 04Storyboard placeholder
Vertical social cuts
Nine-by-sixteen is one of the accepted ratios, and the clip length range covers the short-form window without padding. Write the action so it survives a narrow crop and keep the top third clean for overlays.
00:00:08SHOT 05Storyboard placeholder
Landscape and environment loops
Slow environmental motion is where a stable model shows itself: mist, water, dust, and light changes read as continuous over a full clip length, and failures are obvious as soon as the geometry bends.
00:00:10SHOT 06Storyboard placeholder
Client-facing concept previews
A prompt-only pass costs nothing beyond the generated seconds, so several directions can be tested before anything is shot. Treat the results as discussion material rather than as finished assets.
00:00:12SHOT 07Storyboard placeholder
Storyboard motion tests
Take an approved frame from a board and give it one movement. The result answers a narrow question — does this beat work in motion — without committing to the wider sequence.
00:00:14SHOT 08Storyboard placeholder
Short-form clip extension in post
Because output arrives with an audio track by default, a generated clip can go straight into a rough cut. Request silence explicitly when the audio will be replaced later.
05 / Prompt lab
Prompt starters for each documented workflow
Six starters written for the two routes this page opens. Each one can be loaded into the workspace above and edited in place; treat them as structures rather than as guarantees about the output.
SCENE 01text to video
Text-to-video: a generated establishing shot
No source image. Name the subject, one action, one camera move, and one lighting condition, then state what must stay stable.
View prompt＋
A rain-slicked night market alley, steam rising off the griddles, one vendor looking up as the camera dollies slowly forward through the crowd. Wet asphalt reflections, warm stall lights against a cool sky, steady anatomy, no text and no logos.
SCENE 02text to video
Text-to-video: a landscape with slow motion
Useful when the test is whether a single environment holds together across the full clip length without warping.
View prompt＋
A wide valley at first light, low mist moving between the ridges, one hawk circling high in the frame, camera holding almost still with a very slow tilt upward. Natural colour, no lens flare, no added grain, motion stays believable across the whole shot.
SCENE 03image to video
Image-to-video: portrait with a single small move
The safest way to test a source frame. One action, one direction, and an explicit instruction to leave the background alone.
View prompt＋
Animate the supplied portrait: the subject turns their head slowly toward the lens, blinks once, and a thin strand of hair drifts across the cheek. Keep the background, the clothing, and the lighting direction exactly as they are in the source image.
SCENE 04image to video
Image-to-video: product orbit
Product work depends on the label surviving the camera move, so say that explicitly rather than hoping for it.
View prompt＋
From the supplied product photograph, a slow orbit around the bottle while the label stays legible throughout and the reflection on the table shifts with the camera. Do not redraw the label artwork, do not change the casing proportions, and keep the surface grain unchanged.
SCENE 05image to video
Image-to-video: vertical cut for mobile
Same source frame, different framing intent. Set the ratio first and describe motion that survives a narrow crop.
View prompt＋
Using the supplied photograph as the first frame, a vertical composition: the subject takes one slow step toward the camera as the wind lifts the fabric of their coat. Keep the face and the garment design unchanged, hold the horizon level, and leave the upper third of the frame free of detail.
SCENE 06image to video
Repair prompt for a warped result
When a clip bends at the joints, reduce the scope instead of adding adjectives. Drop the camera move and keep one action.
View prompt＋
Same subject and same setting as before, but with the camera locked off and no zoom. The subject makes one small movement only, and the limbs stay attached and correctly proportioned for the entire clip. Background elements remain static.
COMPARE
Grok Imagine Video 1.5 next to the classic grok-imagine-video
Both columns come from the two official model cards and the vendor model list, checked on 13 September 2026. Prices are the published per-second rates and exclude input charges unless stated.
Grok Imagine Video 1.5 next to the classic grok-imagine-video
Item
grok-imagine-video-1.5
grok-imagine-video
Input modalities
Text, image, and audio
Text, image, and video
Native 1080p
Yes, for text-to-video and image-to-video
No; 480p and 720p only
Published per-second rate
0.08 USD at 480p, 0.14 USD at 720p, 0.25 USD at 1080p
0.05 USD at 480p, 0.07 USD at 720p
Image input price
0.01 USD per image
0.002 USD per image
Video as an input
Not an input modality
0.01 USD per second
Last-frame pinning
Supported through last_frame
Rejected
Combining an image with references
Allowed; becomes reference-to-video with a pinned first frame
Rejected
Published regions
us-east-1 and us-west-2
us-east-1 on the model card, with Batch listed for us-east-1 and us-west-2
Batch API
Supported, with batch pricing listed as not applicable
Supported, with batch pricing listed as not applicable
LIVE PRICING
What Grok Imagine Video 1.5 costs per second
The official model card prices output by the second, and the rate depends on the resolution you ask for: 0.08 USD per second at 480p, 0.14 USD at 720p, and 0.25 USD at 1080p, all checked on 13 September 2026. Billing is described as applying to each second of video generated, and inputs are charged separately: 0.01 USD per image, text input at no charge, and preset-voice audio at no charge. The model list summarises the same model at 0.080 USD per second, which matches the 480p column. Batch pricing is listed as not applicable, and the card notes the model runs on more than one cluster with regional pricing available separately. No promotional period or dated discount appears on the official pages for this model, so the figures above are list rates and not a limited-time offer. This site charges platform points instead of dollars, so the estimate shown in the workspace is the price that applies here, and the official numbers above are the reference the points are set against.
The workspace quotes points before a task is submitted. Points are set by this site and are not the official per-second rate; treat the two as separate scales.
02
Resolution and clip length
The official model ID accepts 480p, 720p, and 1080p, from 1 to 15 seconds with a default of 8. This workspace exposes 720p and 1080p over 2 to 15 seconds with a default of 5, so the cheapest official tier and the one-second floor are both outside what this page can request.
06 / Workflow
From prompt to clip, step by step
01
Choose the route
Pick text-to-video when the idea only exists as words, and image-to-video when a specific frame has to stay intact. The rest of the request is shaped by that choice.
02
Write one subject, one action, one camera move
Name who or what is in frame, what it does, and how the camera behaves. Add the lighting condition and the stability requirement. Two competing actions in a short clip is the most common cause of a warped result.
03
Set the length, quality, and ratio before you submit
Length, quality, and ratio all feed the point estimate, so decide them together rather than adjusting afterwards. If the target is a mobile feed, choose the vertical ratio at this step.
04
Read the estimate, then submit
The workspace shows a point estimate for the current settings. Submit only after you have compared that number against the official per-second rate for the resolution you selected.
05
Review the clip before it leaves the workspace
Check the face and hands, the attachment between subject and props, camera continuity, background structure, and any lettering in frame. If a take is close, change one variable at a time instead of rewriting the whole prompt.
Review before publishing
Grok Imagine Video 1.5 boundaries before you submit
Generation is asynchronous and the wall-clock time is not fixed. The vendor names four factors that extend it: prompt complexity, clip length, resolution, and whether editing is part of the request. Higher resolutions take longer than lower ones, and the request has to be polled until it completes, so a long queue is a scheduling fact rather than a failure of the prompt.
Output is probabilistic. Complex interactions, fast movement, small text, and fine product geometry may need several attempts and a human pass. Different requests built from the same prompt can return materially different clips, and a setting that worked once is not a guarantee for the next submission.
Several documented capabilities are outside what this page opens. Reference images, preset voices, and last-frame pinning are documented for the model ID but are not controls here, and the workspace offers no 480p option and no one-second clip length. Any of these figures read from the vendor documentation describes the model, not this page.
The published figures are list rates without a promotional window, and vendor pricing and regional availability can change. Confirm the current rate on the official model card before making a budgeting decision, and confirm the current controls in the workspace before submitting a task. Commercial use also depends on the vendor terms, the rights attached to any source material, and the rules of the channel you publish to; generating a clip does not by itself settle any of those questions.
Sources
Grok documentation behind this page
Every figure on this page is taken from one of these vendor pages and carries the check date 13 September 2026. Pricing, regional availability, and capability notes can change, so confirm the current values at the source before relying on them.
What is the official model ID for Grok Imagine Video 1.5?＋
The vendor registers it as grok-imagine-video-1.5, with the aliases grok-imagine-video-1.5-preview and grok-imagine-video-1.5-2026-05-30. The hyphens in this page URL are the platform route, not the model ID.
Which workflows does the model ID actually support?＋
Four: text-to-video from a prompt alone, image-to-video from one source frame, reference-to-video using reference images or preset voices, and first-and-last-frame pinning through the last_frame field. Editing and extension are separate endpoints that the vendor documents against the classic grok-imagine-video model.
Does Grok Imagine Video 1.5 support 1080p?＋
Yes, for text-to-video and image-to-video. The documentation states that reference-to-video is capped at 720p. The three priced resolutions are 480p, 720p, and 1080p, and 480p is the documented default.
How long can a clip be?＋
The API accepts an integer duration from 1 to 15 seconds with a default of 8. This workspace accepts 2 to 15 seconds and defaults to 5, so the shortest official option and the official default are both outside what this page can request.
Does the output include sound?＋
Yes by default. The vendor describes sound effects, ambience, and dialogue as generated in the same pass as the picture, and passing generate_audio=False requests a silent video instead. This workspace does not expose that switch.
Can I pin the first and last frame?＋
The model ID supports it: last_frame pins the closing frame, and pairing it with an image pins both ends and lets the model interpolate between them. The vendor notes that this model accepts the combination while the classic model rejects it. No last-frame control is exposed on this page.
Do reference images and voices work here?＋
Reference-to-video is documented for the model ID, with up to three preset voices per request and reference images each holding one element such as a face, a product, or a location. This page does not open a reference control, so treat it as a vendor capability rather than something you can run in the workspace above.
How is the price calculated?＋
By the second, on the resolution requested: 0.08 USD per second at 480p, 0.14 USD at 720p, and 0.25 USD at 1080p, as published on the official model card and checked on 13 September 2026. Image inputs are charged at 0.01 USD each, while text input and preset-voice audio carry no charge. This site bills platform points instead, so the workspace estimate is the figure that applies here.
Why does the workspace show a different length and quality set from the official table?＋
Because the workspace is a product surface, not a mirror of the API. It opens 720p and 1080p over 2 to 15 seconds, while the vendor accepts 480p through 1080p over 1 to 15 seconds. Both sets are correct for what they describe, and the section above sets them side by side so they are not confused.
Is Grok Imagine Video 1.5 the same as the Fast variant?＋
No. The Fast variant is a consumer-side release on the vendor application, described in the launch announcement as producing a six-second 720p clip in about 25 seconds against 40 or more seconds previously. It is not registered as an API model ID, and it is not what this page runs.
Is this the official xAI website?＋
No. This is an independent multi-model creation platform implemented through the official API, and it is not affiliated with, operated by, or endorsed by xAI. The same statement appears at the foot of this page.