Magic Hour Image to Video

Overview

Magic Hour Image to Video turns a photo into a short animated clip through Magic Hour's creation MCP or API. Instead of firing straight at the video model, it treats the job as two steps: get the still frame right, then direct motion on top of it. That order matters because a bad start frame (wrong crop, warped face, altered logo) wastes a paid video generation trying to fix what was already broken in the image.

How it works

1. Define the shot. The agent extracts aspect ratio, subject, one primary action, one camera move, and the details that must stay fixed (identity, product geometry, logos, text, clothing).
2. Build or approve the start frame. If no usable image exists, it generates one; if an image needs a fix (composition, background, cleanup) it edits it first; only a frame that already satisfies the shot skips this step. The agent reviews the still before paying for motion.
3. Animate the approved frame. It writes a motion prompt around what changes over time (subject motion, camera motion, environmental motion) and what must stay fixed, then submits the video job once and waits for it to complete.
4. Inspect and deliver. It checks the first frame, midpoint, and last frame against the approved still, confirms identity/logo/text held up and motion matches the request, then hands back the finished file with the image and video project IDs.

Examples

A typical brief handed to the agent:

Put me in a midnight train station with golden fish swimming through the air. Keep my face and clothes. Fix the still with AI Image Editor if needed, inspect it, then animate my reaction and the fish. Deliver a finished vertical video within my budget.

The agent resolves this into an image-edit call (if needed), a wait_for_image_project check, an image_to_video_create_video call with an explicit motion prompt, a wait_for_video_project check, and a final visual QA pass before returning the video link.

Part of a larger collection

This skill lives inside magichourhq/skills, a monorepo of about 18 Magic Hour skills covering the rest of the image/video/audio workflow: face swap, text-to-video, image editing, talking/lip-sync video, character replace, video editing, music video, body swap, character consistency, thumbnails, product visuals, campaign kits, brand kits, marketplace images, and explainer video. They all share the same Magic Hour MCP connection, so installing one doesn't require installing the rest. Browse the full list at github.com/magichourhq/skills.

Installation

You need a Magic Hour account and API key, plus an agent client that supports skills (Claude Code, Codex, or another skills.sh-compatible client).

1. Create a Magic Hour account and API key at magichour.ai if you don't have one already.
2. From your project root, run npx skills add magichourhq/skills --skill magic-hour-image-to-video to install just this skill.
3. Connect Magic Hour's creation MCP at https://mcp.magichour.ai/ in your agent client, using your API key.
4. Verify it worked by asking your agent to check the connection (an authenticated account read) before generating anything.
5. Attach a photo and describe the scene/motion you want; the skill will handle image-fix, animation, and inspection steps automatically, spending credits according to your stated budget.