How to do the zombie AI trend

Updated 2026-10-08

Short answer

You need two photos: one of the person who turns into a zombie and one of the person who can't shoot. Give both to an image-to-video AI with a four-beat prompt — the turn, the gun, a warm memory, the gun lowered — in 9:16 for about 15 seconds. Our zombie AI video generator does it in one click; Dreamina, CapCut or Google Flow can do it with your own prompts and some editing.

The four beats

Time	Beat	What happens	Light
0–3 s	The turn	Image 1 appears as a zombie, still recognizable	Cold
3–6 s	The gun	Image 2 raises a gun, shaking, can't fire	Cold
6–10 s	The memory	Both, healthy and happy, somewhere warm	Warm
10–15 s	The embrace	Back in the dark, the gun drops, they hug	Cold

Method 1 — one click

Add both photos to the generator, pick a scene (abandoned house, snowed-in cabin, gas station at night or hospital corridor) and your bond (couple, parent and child, or best friends), and press Generate. You get one finished 15-second 9:16 video with sound — no prompt, no editing. It runs on MiniMax H3 with both photos as references, so the faces hold from the first beat to the last.

Method 2 — Dreamina or CapCut

Dreamina and CapCut's AI video tools run ByteDance's Seedance models. The usual approach takes two generations:

Start an image-to-video generation that accepts reference photos and add both photos.
Generate the zombie scene with beats 1, 2 and 4 of the prompt below.
Generate the memory scene separately with beat 3.
Join the clips in CapCut, put the memory in the middle, and add a soft piano track at the cut.

Method 3 — Google Flow or Gemini (Veo)

Add both photos as references (ingredients) for the video.
Set the aspect ratio to 9:16.
Paste the full prompt below. If the tool names photos differently, replace "Image 1" and "Image 2" with "the first photo" and "the second photo".

Copy-paste prompt

This is the exact prompt behind the abandoned-house scene. Image 1 is the person who turns; Image 2 is the one who can't shoot.

Image 1 is the person who has turned. Image 2 is the survivor. Keep both faces, hairstyles and builds exactly as in their photos and never swap them. One continuous 15-second cinematic vertical film. No text, no subtitles, no other people.
0–3s: A dim abandoned house at dusk during a zombie apocalypse, cold blue-grey light. The person from Image 1 stands in the doorway as a zombie — ashen grey skin, clouded white eyes, torn dusty clothes — still clearly recognizable, and staggers forward. Low drone, creaking floorboards.
3–6s: Close-up of the person from Image 2 raising a pistol with shaking hands, tears running down their face, breathing hard, unable to pull the trigger.
6–10s: Match cut to a warm golden-hour memory: the two of them, healthy and exactly as in their photos, laughing together in a sunlit wheat field. On the first warm frame the music shifts to soft piano and warm strings.
10–15s: Back in the dark room, the person from Image 2 slowly lowers the gun and opens their arms; the zombie stops and rests its head on their shoulder in an embrace. Piano continues, a quiet sob.
35mm film grain, shallow depth of field, slow handheld camera.

When it goes wrong

The faces swap or drift: Say which photo plays which role in the first sentence, and use photos with one face each, facing the camera.
The zombie doesn't look like them: Keep "still clearly recognizable" in the prompt and use a sharp, well-lit face photo for Image 1.
The memory doesn't feel warm: Name the light change outright: cold blue-grey before, golden hour in the memory, and the music shift on the first warm frame.
The tool refuses the gun: Some tools filter weapons. Try "raises a flare gun" or drop the weapon: "raises a hand to stop them, then lowers it".

Skip the prompts and the editing.

15 s, 9:16, sound included, both faces kept.

Make my zombie video