Grok Imagine Video Edit
What is Grok Imagine Video Edit?
Grok Imagine Video Edit is xAI's video-to-video editing endpoint. Pass a source clip and a single-sentence prompt and the model returns an edited version of that clip with the same duration and aspect ratio. Style transfers, atmosphere changes, and prop swaps are the strongest cases. The input cap is 8.7 seconds, so this is a tool for shot-level fixes rather than full edits, and the output resolution matches the input. There are no duration, aspect_ratio, or resolution knobs to set: the endpoint preserves the source. The practical use is the late note: a client wants the cup to be a wine glass, the sky should be sunset instead of midday, the look should be more cinematic. One API call, a 60-second wait, and the cut is back in review.
Key features of Grok Imagine Video Edit
Four features cover the surface area you'll use in a real edit pass.
Single-sentence edits on existing footage
Send `prompt` and `video_url`. The model rewrites the clip according to the instruction. No keyframes, no rotoscoping, no compositing graph.
Source-preserving output
The output keeps the source duration, aspect ratio, and resolution. No conversion step, no aspect re-fit on the way back into the timeline.
Style transfer, atmosphere, and prop swaps
The strongest edit categories. "Make it 1970s film", "turn this into a rainy night", "swap the cup for a wine glass" are the kind of one-line briefs that come back cleanly.
8.7 second input cap
Maximum source clip is 8.7 seconds, which fits a typical hero shot, ad cutaway, or social cut. Trim long takes before sending.
Best for
Late-stage creative pickups
Client wants a different prop or colour grade. Edit one shot in 60 seconds instead of reshooting or re-rendering the whole timeline.
Style transfer for social repurposing
Take a brand shoot and restyle it into "vintage film", "anime cel", or "noir" for platform-specific cuts. Source duration and aspect stay locked.
Atmosphere changes
Day to night, sunny to rainy, calm to stormy. Useful for stock footage that needs a different mood without reshooting.
Colour grading without DaVinci
Prompt-based colour treatment. Faster than building a node tree for a one-off social cut.
Prop substitutions
Swap a coffee cup for a wine glass, a phone for a tablet. Works best when the new prop is similar in shape and position to the original.
Localization tweaks
Change signage, labels, or background text without reshooting. Useful for ads that run in multiple regions from one master.
Use cases
Handle late client notes on a finished cut by sending each shot through one prompt. The cup becomes a wine glass, the sky becomes sunset, the look becomes more cinematic, all in one API call per shot. Repurpose a brand shoot for a platform-specific cut by restyling each clip with a single sentence ("vintage 1970s film", "anime cel"). Localize an ad set by changing signage, labels, or background text per region without reshooting. Add a one-click "warmer grade" or "rainy night" preset to a video editor that maps each user choice to a fixed prompt.
Limitations
The 8.7-second input cap means this is a tool for shot-level fixes, not edit-length passes. Output resolution matches the source. Major geometry changes (different actor, new camera move) are less reliable than style and grade work. There is no aspect-ratio override: the output retains the source aspect ratio, so reframing has to happen in your NLE before or after the call.
API examples
Call Grok Imagine Video Edit from any language by POSTing to /v1/tasks. Full parameter docs live at docs.unifically.com/models/video/xai/grok-imagine-video-edit.
curl -X POST https://api.unifically.com/v1/tasks \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{
"model": "xai/grok-imagine-video-edit",
"input": {
"prompt": "Make the sky turn purple and add shooting stars",
"video_url": "https://example.com/clip.mp4"
}
}'
Successful submission returns a task_id. Poll GET /v1/tasks/<task_id> or set a callback_url on the request to receive the finished video URL.
FAQs
People also ask
It takes a video clip and a text prompt and returns an edited version of that clip. Restyle the look, change the time of day, swap a prop, or adjust the mood from a single sentence. No NLE or rotoscoping required.
Two required fields. prompt is the edit instruction in plain language. video_url is the source clip. The maximum input length is 8.7 seconds; the output keeps the source duration, aspect ratio, and resolution.
No. This is video-to-video only. For a clip from a prompt alone, use the regular Grok Imagine endpoint. Use this one when you already have footage and need to alter it.
Up to 8.7 seconds. Past that, trim before sending. Output preserves the source duration, so a 6-second input returns a 6-second edited clip.
Typically 30 to 90 seconds depending on source length and edit complexity. Poll GET /v1/tasks/<task_id> until the status moves to completed, or pass a callback_url and wait for the webhook.
Style transfers ("make it look like a 1970s film"), atmosphere changes ("turn this into a rainy night"), colour grades ("warmer, more golden hour"), and prop substitutions ("replace the cup with a wine glass") are reliable. Major geometry changes (different actor, new camera move) are less so.
No. The output retains the source duration and aspect ratio, with resolution matched to the input. The duration, aspect_ratio, and resolution parameters are not supported on this endpoint.