Understand Grok Imagine Video 1.5 before you build on it.
Grok Imagine Video 1.5 is SpaceXAI's image-to-video model, published on 16 June 2026 and out of preview in the vendor's Imagine API as grok-imagine-video-1.5. A faster tier, Video 1.5 Fast, shipped to Grok Imagine on the web and to the iOS and Android apps at the same time. The release post sells it on three axes — sound effects, ambience and dialogue generated in the same pass; motion that holds together across a clip; and roughly double the generation speed of the model it replaces.
Every figure on this page comes from SpaceXAI's release post for Grok Imagine Video 1.5, its API documentation for the grok-imagine-video-1.5 model, its video-generation guide, its Imagine pricing table and its developer FAQ. This page carries no gallery, because the vendor publishes no still frames generated by this model: the only images on its release page are screenshots of the Grok Imagine application itself. SpaceXAI publishes no standalone licence or output-ownership document here, and this page does not fill that gap.
How SpaceXAI describes Grok Imagine Video 1.5
The release post is written as a list of improvements against the previous model rather than as a specification sheet, and each claim is scoped that way. Read it as the vendor's account of what changed in 1.5, not as a comparison against anyone else's video model.
Audio generated in the same pass
Sound effects, ambience and dialogue come out with the video rather than being added afterwards, and the post says they land on the action. Speech is described as clearer and better synced. This is the release's headline claim.
Motion and physics that hold
Movement holds together over the length of a clip, with fewer warps and more believable weight and momentum — stated as an improvement over the previous model, not as an absolute.
Speed, with a number attached
Video 1.5 Fast produces 6-second 720p clips in about 25 seconds, against 40+ seconds in the previous model. SpaceXAI describes that as almost doubling generation speed.
Out of preview in the API
1.5 left preview and became generally available as grok-imagine-video-1.5. The documented call is a starting image, a motion prompt, and a choice of resolution and duration.
The workflow changed too
Projects for organising work, several generation agents running in parallel, and search across your library all arrived alongside the model over the following days, which is the part of the release that is not about the model at all.
What the release post and the API docs document
Only the modes and controls SpaceXAI names are listed here. The API documentation is the more precise of the two sources, so the parameter values below come from it.
Image-to-video
The mode the release is built around: give the model a starting image, describe the motion, and choose a resolution and a duration. Output defaults to the input image's aspect ratio unless you override it.
Text-to-video
A prompt alone, with no starting frame. This is one of the two modes where 1080p is available; the prompt is required here and optional in the modes that carry an image.
Reference-to-video with pinned frames
Reference images steer the output without forcing the first frame. On 1.5 a starting image can be combined with references, or a last frame can be pinned on its own so the model generates the opening and lands on it. The older grok-imagine-video rejects all of that, and reference-to-video is capped at 720p.
Reference audio and preset voices
1.5 can carry a voice through reference_audios, drawn from the vendor's built-in roster and named by voice_id. Using your own audio files is open to trusted partners on request rather than to every account.
Video editing
An existing clip can be edited, but with two hard limits the docs state outright: you cannot set a duration or an aspect ratio. The output keeps the original's both, and the duration it keeps is capped at 8.7 seconds.
Clip configuration
Duration runs 1 to 15 seconds, resolution is 1080p, 720p or 480p with 480p as the default, and the ratio table covers 1:1, 16:9, 9:16, 4:3, 3:4, 3:2 and 2:3 with 16:9 as the default.
Documented specifications
Every row below is stated in SpaceXAI's release post, its API documentation, its video-generation guide, its pricing table or its developer FAQ. Rows marked as unpublished are gaps the vendor has not filled, not values this page has estimated.
- Developer
- SpaceXAI
- Model
- Grok Imagine Video 1.5
- Release date
- 16 June 2026
- API model id
- grok-imagine-video-1.5
- Inputs
- A text prompt, a starting image, reference images and/or reference audio
- Clip duration
- 1 to 15 seconds
- Edited clip duration
- Keeps the original and cannot be set; capped at 8.7 seconds
- Resolutions
- 1080p, 720p or 480p, with 480p as the default
- Resolution limit
- Reference-to-video is capped at 720p
- Aspect ratios
- 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3; 16:9 is the default
- Audio
- Sound effects, ambience and dialogue generated in the same pass
- Voice references
- Built-in roster named by voice_id; your own audio files on request to trusted partners
- Fast tier
- Video 1.5 Fast: about 25 seconds for a 6-second 720p clip, against 40+ seconds previously
- API price
- $0.080 per second; the previous grok-imagine-video is $0.050 per second
- Watermark
- A Grok mark is applied to output; the vendor states there is no setting to remove it
- Licence
- Not published
- Output ownership
- Not published
Documented access channels
SpaceXAI ships this model in two places: its own consumer product, where the fast tier landed, and its API, where 1.5 came out of preview. The vendor publishes a per-second rate for the second and nothing comparable for the first.
Grok Imagine on the web
Where Video 1.5 Fast rolled out, alongside the existing image models. SpaceXAI does not publish which tier the app serves by default.
OpenThe model page
The vendor's own page for grok-imagine-video-1.5, carrying its per-second rate, its resolution options and its regional pricing note.
OpenVideo generation guide
The endpoint documentation, including the duration range, the ratio table, the resolution limits and the modes that accept reference audio.
OpenImagine pricing
The vendor's rate table for every Imagine model, still listing the previous grok-imagine-video beside its replacement.
OpenThe release post
SpaceXAI's own announcement for this model, and the page every claim on this page about what changed in 1.5 is drawn from.
OpenGrok Imagine Video 1.5 FAQ
Does Grok Imagine Video 1.5 generate its own audio?
Yes. SpaceXAI states that sound effects, ambience and dialogue are generated in the same pass as the picture and land on the action, with speech clearer and better synced than in the previous model. There is also a reference-audio mode that carries a preset voice, drawn from the vendor's own roster and named by voice_id.
How long can a clip be?
The documented range is 1 to 15 seconds. Editing an existing clip is the exception: you cannot set a duration there, and the output keeps the original's, which is capped at 8.7 seconds. The same restriction applies to aspect ratio when editing.
Can I get 1080p output?
Yes, for text-to-video and image-to-video. Reference-to-video is capped at 720p, and 480p is the default whenever you do not specify a resolution. The three documented options are 1080p, 720p and 480p.
What does it cost?
$0.080 per second of generated video for grok-imagine-video-1.5. The previous grok-imagine-video is listed at $0.050 per second. SpaceXAI prices video per second, and notes that both duration and resolution affect the total.
Is this xAI's model or SpaceXAI's?
Both names appear on the vendor's own pages. The site and the model pages are branded SpaceXAI, the footer reads © 2026 SpaceXAI LLC, and the release post's sibling benchmark note says its models are listed on Arena under SpaceXAI. The same site's news item dated 2 February 2026 is headed xAI joins SpaceX and records SpaceX announcing it had acquired xAI, and the developer documentation still uses xAI in places. This page treats SpaceXAI as the publisher because that is the name the product pages currently carry.
Can I use the output commercially, and does it carry a watermark?
SpaceXAI publishes no output-ownership or licence statement for this model, so this page cannot answer the first half. What governs your use is the terms attached to the channel you reach the model through, together with the vendor's Acceptable Use Policy. On the second half the answer is definite: generated videos carry a Grok watermark, there is no setting to remove it, and the Acceptable Use Policy prohibits removing or obscuring it.