Midjourney video starts from an image. Choose an existing Midjourney image or upload a starting frame, then use automatic animation or write a manual motion prompt. Direct what changes over time instead of redescribing everything already visible.

Midjourney's official video documentation explains the current workflow, low- and high-motion choices, extensions, and raw MP4 downloads.

Formula: subject action + camera move + environmental motion + timing + preservation constraints.

Low motion or high motion?

Low motion suits portraits, products, and stable scenes. High motion suits dynamic action, but gives the model more opportunities to alter the source. Start low, then increase only when necessary.

12 copy-ready prompts

1. Subtle portrait

The subject takes one quiet breath and glances slightly toward the window. Hair moves gently. Locked camera. Preserve facial identity, skin texture, clothing, and background. No talking, morphing, or exaggerated expression.

2. Slow push-in

The camera performs a very slow push-in from medium close-up to close-up. The subject remains still. Preserve face, pose, wardrobe, color palette, and light direction. No cuts or sudden acceleration.

3. Product highlight

The product remains perfectly still as the camera pushes in slowly. A soft highlight travels across the surface. Preserve exact shape, logo, typography, label, color, reflections, and packaging.

4. Turntable shot

The product rotates clockwise by 20 degrees while the camera stays locked. Preserve package geometry, branding, label placement, cap, and scale. No full spin or text mutation.

5. Fashion tracking shot

The subject walks forward while the camera tracks sideways at eye level. Fabric and background parallax move naturally. Preserve identity, outfit, proportions, and location.

6. Controlled orbit

The camera makes a slow 30-degree clockwise orbit around the still subject. Preserve facial identity, body shape, lighting, clothing, and environment. No full rotation.

7. Landscape glide

The camera glides slowly forward above the path. Grass, leaves, and clouds move gently. Preserve terrain, horizon, buildings, and season. No time-lapse flicker.

8. Rainy window

Rain droplets travel down the foreground glass while the room remains stable. Add realistic refraction and a slow push-in. Preserve the person, furniture, frame, and lighting.

9. Food steam

Thin steam rises from the dish while the camera moves closer. Preserve food arrangement, garnish, plate, table, and lighting. No ingredient mutation or excessive smoke.

10. Magical particles

A small trail of golden particles follows the subject's hand for two seconds and fades. Preserve identity, anatomy, environment, and lighting. No transformation.

11. Vertical reveal

In 9:16, the camera pulls back slowly to reveal the full outfit and environment. Preserve face, clothing, hands, background geometry, and palette. No jump cut.

12. Day to evening

Daylight shifts gradually toward early evening while the camera stays locked. Interior lights turn on softly. Preserve buildings, landscape, weather, and composition.

Testing workflow

  1. Start with a sharp, coherent image.
  2. Choose low motion first.
  3. Ask for one action and one camera move.
  4. List only essential preservation constraints.
  5. Extend only an approved clip.
  6. If it fails repeatedly, test another model with the same source.

Check the still with the consistency checker, draft motion with the cinematic prompt generator, or compare workflows in Video Lab.

Midjourney video FAQ

How do prompts work?

Start with an image and describe how its subject, camera, and environment should move.

Low or high motion?

Start low for stability. Use high motion when stronger movement is essential.

Is it text-to-video?

The documented workflow begins from a starting image, which can itself be generated from text.

How do I reduce warping?

Use a clean source, one action, one move, lower motion, and explicit preservation constraints.

Related guides