Models / Grok Imagine / Video to video

Grok Imagine video to video prompting guide

How to prompt Grok Imagine in video-to-video mode. Every point links its source; community tips are marked. Your agent gets the same guide through the Atlas MCP (get_guide).

Prompt structure

edit instruction→what stays the same

• No dedicated prompting guide published; describe subject, camera, pacing and sound in plain language. source ↗

What the text should describe

• /v1/videos/edits on grok-imagine-video: source .mp4 up to 8.7 s; output keeps source duration/aspect, capped at 720p. The model applies the change and keeps the rest — 'say what must stay the same'. Examples: 'Give the woman a silver necklace', 'Change the color of the woman's outfit to red'. source ↗

What not to describe

• duration, aspect_ratio and resolution are not supported on edits; one change per request works best (xAI runs separate edits concurrently). source ↗

Responds well to

• t2v first generates a frame from your prompt then animates it, so the opening frame description matters. source ↗

Duration and limits

• Duration 1–15 s (default 8). 1.5 native 1080p for t2v/i2v; r2v up to 720p. i2v follows the image's aspect ratio. Editing keeps source length (max 8.7 s), 720p. source ↗

Negative prompt

Supported via .

Real prompts, credited

Give the woman a silver necklace / Change the color of the woman's outfit to red / Give the woman a wide-brimmed black hat

Next

Sources

• docs.x.ai/developers/model-capabilities/video/editing
• docs.x.ai/developers/model-capabilities/video/generation