◈ ZombieAI

Guide · The trend

The zombie trend prompt, shot by shot

By the Zombie AI team · Updated 2026-10-20

Two reference prompts: a 14-second couple version and a 15-second pet version.

What each part does

PartWhat it does
Opening lineSets the duration, format, camera, grain, audio and “no dialogue” in one sentence.
Face lockAssigns Photo 1 and Photo 2 and forbids swapping or merging the two identities.
Present sceneSets the cold room, the gun, the infected look and the moment of recognition.
TimestampsGives each beat a time so the model cannot skip the ending or compress the memory.
TransitionA hard cut for the couple, a cross-dissolve for the pet; both keep the same pose and framing.
MemoryWarm light, healthy faces, the same two people, a held last frame.
Camera / hugKeeps the camera fixed and the embrace readable: one head and two arms per person, no extra limbs.
ConstraintsNo gunshot, no gore, no slow motion — the story stays romantic, not a horror payoff.

A prompt like this works because it removes choices. It fixes the two identities, the order they appear in, the cold-to-warm transition and the final pose. The timestamps stop the model from skipping the memory; the constraints stop it from turning the scene into a gunshot or a gore payoff. The room, the clothes and the light are decoration around those fixed decisions.

Couple prompt (14 seconds)

Photorealistic cinematic zombie-horror, 14 seconds, live-action, 35mm grain, handheld camera, native audio, no dialogue.
Faces locked to two reference photos: Photo 1 is the survivor, Photo 2 is the infected lover. Never swap or merge the two identities.
Present — abandoned farmhouse kitchen, cold blue-grey overcast light. The survivor (Photo 1), torn olive-green anorak, dried blood on the forehead, tears, trembles while aiming a black pistol at their infected lover (Photo 2): pale dirty skin, milky-white eyes, dark veins, torn clothing. The lover snarls and screams at them. The survivor cannot shoot — lowers the pistol, sets it on the floorboards, opens their arms. The lover screams again and lunges, grabbing the survivor's face. Hard smash cut at 9.5s, the snarl cut off instantly.
Memory — warm golden-hour wheat field, healthy clear-eyed faces, laughing embrace, hand-in-hand walk, then sitting together beneath the sunset as the camera pulls back into warm flare. Hold the last frame.
No gunshot, no gore, no slow motion.

Photo 1 is the survivor; Photo 2 is the infected lover. Use a model that accepts two reference images and a 14-second output. Zombie AI itself uses a fixed 15-second script and has no prompt box.

Pet prompt (15 seconds, vertical)

Photorealistic cinematic zombie-horror turning into warm romance, vertical 9:16, 15 seconds, live-action, 35mm grain, slow drifting camera, shallow depth of field, native audio, no dialogue.

Locked to two reference photos: Photo 1 is the survivor, Photo 2 is the pet — match its species, size, markings and face exactly. Never swap or merge them. In the present the pet is infected — dull matted coat, clouded eyes, a raw wound on the body, head low. In the memory the same pet is healthy and glossy. One readable face per shot before the transition.

Present, 0:00–0:07.6 — cold teal farmhouse kitchen, bright window with bare branches.
1. 0:00.0–0:01.5 — over-the-shoulder: the survivor's dark out-of-focus back in the near foreground, pistol aimed; the infected pet stands in focus in the mid-ground, backlit by the window. Camera level with the pet.
2. 0:01.5–0:02.8 — cut to close-up, survivor's face, tears and blood, pistol aimed into the lens.
3. 0:02.8–0:04.3 — cut to close-up of the infected pet, taking a slow stiff step toward camera. Loudest sound here: a low wet animal sound rising into a snarl.
4. 0:04.3–0:07.6 — cut to close-up, survivor, ONE continuous take: the pistol lowers out of frame, the face softens, they lower toward the pet, an open palm rises into frame, the pet comes to the hand, and the survivor wraps both arms around it, crying into its coat. Hold exactly on that embrace.

CROSS-DISSOLVE at 0:07.6, about 0.2s — NOT a cut, fast, with identical pose and framing on both sides. The music hits at full level on the dissolve, not a fade-in. No hard cut, no flash, no jump.

Memory, 0:07.6–0:15.0 — warm golden-hour autumn forest, the pet now healthy.
5. 0:07.6–0:09.8 — same pose: the survivor holds the pet, crying and smiling, sunbeams and lens flare behind them.
6. 0:09.8–0:11.6 — the survivor straightens up, strokes the pet, laughs down at it, both in a sunbeam.
7. 0:11.6–0:13.1 — the survivor crouches, straightens, and turns to the path, the pet moving up beside them.
8. 0:13.1–0:15.0 — wide, single-point perspective down the forest path: the two move away from camera silhouetted against the low sun, camera pulling back until they are small in the lower third. Hold.

No gunshot, no gore, no slow motion.

Photo 1 is the survivor; Photo 2 is the pet. Use a model that accepts two reference images, a 15-second output and a vertical frame.

If the result drifts

What you seeWhat to change
The faces swap or mergeKeep the role assignment and the “never swap or merge” line; use two separate photos.
The zombie bites instead of huggingAdd “instead of attacking, they hug”; use the hug as the end frame if the model supports it.
The face changes across the transitionUse the same reference photo and repeat one fixed detail — necklace, collar, hoodie, fur marking.
Extra hands or a duplicated headAsk for a fixed side-on camera and “one head, two arms” per person.
The memory looks like a different personSay “same face, same detail, alive and healthy”; match the pose and framing.

FAQ

Is this the prompt Zombie AI uses?

No. Zombie AI uses a fixed 15-second script and has no prompt box. These are reference prompts for another image-to-video model. The story beats are the same; the timing and staging are written out here.

Which prompt should I use?

Use the couple prompt for two people and a 14-second output. Use the pet prompt for one person and one dog or cat, with a 15-second vertical output. Do not mix the cast: Photo 2 must match the prompt’s role.

Why does the pet version dissolve instead of cutting?

The pet version is staged as a slower, warmer transition: the survivor is already holding the animal, so a 0.2-second cross-dissolve keeps the same pose and lets the music land. The couple version uses a hard cut at the lunge.

Can I change the setting?

Yes. Keep the parts table intact: two identities, timestamps, the transition, the memory and the constraints. Change the room, the clothes or the memory location, not the role assignment or the cut.

Is the wound or dried blood gore?

No. They are single details, not a gore scene. The prompt still says no gunshot, no gore and no slow motion; the story stays focused on the recognition and the hug.

Make your zombie video

Two photos, 15 seconds, with sound and no watermark.

Make your video

More guides