Video GenerationActive
xai / grok-imagine-video-extension
grok-imagine-video-extension
Edit or extend a short clip with xAI Grok Imagine. Upload a source video and a detailed prompt to generate a continued MP4 sequence.
Image to Video
Model ID
grok-imagine-video-extension
Provider
xai
Updated
1776347130
wiro playground—xai/grok-imagine-video-extension
Updated 1776347130
## Overview
Grok Imagine Video Extension is a prompt-driven video extension and editing model from xAI. It continues a clip from its last frame or applies guided edits. Use it for short, cinematic continuations with strong scene consistency.
## What you can build
- Extend a clip past its last frame to add a new action beat
- Continue a camera move, like a pan, dolly, or slow zoom
- Add small story moments like a character turning, waving, or picking up an object
- Patch a shot by editing a detail, like color, clothing, or a prop
- Create longer social clips by extending in controlled segments
## Inputs
- **Input Video** (required)
- Type: single video file
- Default: `xai-grok-image-video-extension-input-1.mp4`
- Notes: Use **up to 8.7 seconds** for edit mode. Use **2 to 15 seconds** for extend mode. Only 1 video is allowed.
- **Prompt** (required)
- Type: text
- Default: a long cinematic example prompt is prefilled
- Notes: Describe what should change (edit) or what happens next (extend). Keep prompts within 4096 characters.
- **Video Mode** (required)
- Type: select
- Options: `edit` (Edit Video), `extend` (Extend Video)
- Default: `edit`
- Notes: Edit modifies the clip. Extend continues from the last frame.
- **Extension Duration** (extend mode only)
- Type: number (seconds)
- Range: 2 to 10 (step 1)
- Default: 5
- Notes: Ignored in edit mode.
## Outputs
The model returns 1 video file.
- Format: MP4 (`video/mp4`)
- Frame rate: typically 24 fps
- Resolution: typically 480p or 720p
- Extend mode: output includes the original clip plus the new extension stitched into one MP4
- Edit mode: output keeps the input clip length and matches input resolution, capped at 720p
## Recommended settings
- **Clean continuation (most cases)**
- Video Mode: `extend`
- Extension Duration: 3 to 6 seconds
- Prompt: describe motion, camera movement, and the next action
- **Bigger story beat**
- Video Mode: `extend`
- Extension Duration: 7 to 10 seconds
- Prompt: add timing cues, like “after 2 seconds,” and name key objects to keep stable
- **Precise change to an existing shot**
- Video Mode: `edit`
- Prompt: ask for one change at a time
## Limitations
- Edit mode input video length is capped at **8.7 seconds**.
- Extend mode requires an input video length of **2 to 15 seconds**.
- Extend mode supports **2 to 10 seconds** of new footage per run.
- Input videos must be MP4 files, using common MP4 codecs like H.264, H.265, or AV1.
- Output resolution is limited to **480p or 720p**.
- Content moderation can block a generation even if the prompt looks harmless.
## Safety & compliance
xAI applies content policy checks to generated video.
- Don’t create sexual content involving minors. Never attempt to depict minors in sexual contexts.
- Don’t generate non-consensual intimate imagery or sexualized content of real people.
- Don’t use the model for harassment, doxxing, stalking, or defamation.
- Don’t generate instructions or content that enables wrongdoing or physical harm.
- Respect copyright, trademark, and privacy rights in source videos and prompts.
Example prompts
Great starting points for grok-imagine-video-extension.
Cinematic, smooth tracking shot continuing the slow pan from right to left. As the camera glides past the foreground baskets of lemons, apples, and the large orange pumpkin, a gentle coastal breeze sweeps through the scene. The dense clusters of hanging dried white garlic, red chili peppers, and gourds sway softly and independently with accurate pendulum physics and realistic collision, without morphing or losing their complex, papery textures. From the left edge of the frame, a weathered vendor\'s hand reaches in to gently pick up one of the bright yellow lemons from the woven basket, demonstrating flawless anatomical geometry and precise object interaction. The deep background remains perfectly stable, maintaining the rigid, high-resolution structure of the sunlit, rocky green mountains and the clear blue sky without any warping or hallucination during the continued camera movement. Crisp, realistic shadows cast by the bamboo roof shift naturally across the vibrant produce.Video Generation
Cinematic, slow panning shot continuing smoothly to the right while gently tilting down over the ornate wrought-iron railing, revealing more of the steep drop, the dark blue Mediterranean sea, and the sandy beach far below. The highly complex, clustered geometry of the colorful cliffside houses and the rugged mountain terrain remains perfectly rigid and distinct, with zero morphing, melting, or architectural hallucination. Small, realistic waves roll and crash gently on the distant shoreline with accurate fluid dynamics. The sharp, intricate shadows cast by the railing on the sunlit stone terrace maintain perfect structural tracking and perspective with the camera\'s movement. A small flock of seagulls flies dynamically through the deep midground gorge, displaying flawless avian flight physics and accurate depth perception against the vast, bright landscape.Video Generation
Edit the video so the illuminated spokes and outer rim of the Ferris wheel dynamically and smoothly cycle through a continuous, vibrant RGB color spectrum (shifting seamlessly through neon pink, purple, red, golden yellow, and bright cyan) as it rotates. The color transitions must be fluid and temporally consistent across the rigid metal geometry of the wheel without causing the structural spokes or passenger cabins to warp, melt, or hallucinate. The shifting neon colors must cast physically accurate, dynamically changing ambient light and reflections onto the attached white passenger cabins, the dark pavement below, and the surrounding architectural facades. The twilight sky, background city streetlights, and overall camera movement must remain perfectly stable and completely unaffected by the edit.Video Generation
Edit the video to completely remove all boats, watercraft, and their trailing white water wakes from the Grand Canal. Fill the empty spaces seamlessly with continuous, natural-looking green canal water. The newly generated water surface must feature realistic, gentle ripples, accurate fluid dynamics, and stable, undisturbed reflections of the surrounding historic Venetian buildings and the bright sky. The complex architectural geometry of the palazzos, the distant basilica, the clear daytime lighting, and the overall perspective must remain perfectly intact, without any morphing, hallucination, or structural distortion throughout the shot.Video Generation
API quick start
Run grok-imagine-video-extension with a single API call.
POST https://api.wiro.ai/v1/Run/xai/grok-imagine-video-extension
{
"prompt": "Cinematic, smooth tracking shot continuin…",
"inputVideo": "https://your-cdn.com/input.mp4",
"videoMode": "edit",
"duration": 5
}