AI VideoAI Video · Lesson 04

Image-to-Video Workflow

Lock the frame as a still, then animate — the professional route.

Video tutorialTrack course · Youri van Hofwegen
Speed
Next lesson
Watch 00:00 · 2 checkpoints Watch on YouTube More on this topic
Transcript & captions
10/10

Text-to-video gives you a lottery ticket. Image-to-video gives you a storyboard you already approved. Generate (or shoot) a keyframe, fix it until it is perfect, then hand it to the motion model with a short motion-only prompt.

  • Step 1 — Generate the keyframe at high resolution with an image model.
  • Step 2 — Repair it: faces, hands, text, logos. A flaw in the still becomes a flaw in every frame.
  • Step 3 — Write a motion prompt only: what moves, and how the camera moves.
  • Step 4 — Generate 3-5 variations, keep the one with the least warping.
Motion-only promptjs
// The image already defines subject, environment and grade.
const motion = 'Subject breathes and slowly turns head to camera. ' +
  'Camera: gentle 5% push-in. Hair and steam drift. No cuts.';

Aspect ratio is decided by the keyframe. Produce 16:9 for YouTube and a separate 9:16 keyframe for Shorts/TikTok — never crop a 16:9 generation to vertical.

Knowledge check

0/2 answered

What should an image-to-video prompt describe?

A flaw in the keyframe will...