These are the zombie AI prompts behind the films on this site, rewritten for any image-to-video tool that takes two reference photos: Google Flow (Veo 3.1 "ingredients"), Dreamina, Kling or similar. They are tested, and the notes below explain what broke before they worked.
Upload the photos in this order every time: the survivor first, the one who turns second. The prompts refer to "the first" and "the second reference image".
Reel 1: the standoff
Vertical 9:16 cinematic zombie love story, 35mm film, cold desaturated blue-grey grade, heavy grain. Exactly two characters. The woman from the first reference image is the survivor: keep her exact face, hairstyle and outfit, now dusty and soaked. The man from the second reference image has turned into a zombie: still recognizably the same man with the same hairstyle and outfit, but with ashen grey cracked skin, milky white eyes with no pupils, dark veins on his neck and torn, muddy clothes; no blood, no gore. An empty rain-soaked city street at dusk, abandoned cars, flickering streetlights. 0-2s: wide shot, he staggers slowly toward her. 2-4s: medium close-up from the side: she raises a pistol with both trembling hands and points it at him across the gap, tears streaming, unable to pull the trigger. 4-6s: close-up of his milky white eyes, a flicker of recognition; she lets the pistol fall and opens her arms. 6-8s: he stumbles into her arms and she holds him tight, eyes closed. The pistol is never fired. Audio: steady rain, a distant siren, her shaky breathing, one soft piano note; no dialogue. No text, no watermark.Reel 2: the memory
Vertical 9:16 cinematic flashback, warm golden-hour light, soft halation, gentle film grain. Exactly two people, both healthy and smiling: the woman from the first reference image and the man from the second, exact faces and outfits, no zombie features at all. A quiet sunlit beach at sunset. 0-3s: he runs into her arms and they spin around, laughing. 3-6s: they walk barefoot along the waterline holding hands. 6-8s: close-up, foreheads touching, eyes closed, soft lens flare. Audio: gentle waves, warm laughter, a tender piano melody; no dialogue. No text, no watermark.Join the two clips with a hard cut at the embrace. That is the whole zombie couple video.
Variations
- Best friends: in reel 2, replace the beats with "they crash into a laughing hug, one gives the other a piggyback ride, close-up with arms around each other's shoulders".
- Pet version: describe the second image as "the dog from the second reference image… same breed, coat and collar, but patchy grey fur and milky white eyes". In reel 2: "the healthy dog bounds into his arms, tail wagging; he throws a ball and it races after it". See the zombie dog video.
- Other worlds: a frozen highway that cuts to a summer road trip, a dead fairground that cuts to the fair at night, a ruined chapel that cuts to a first dance.
What breaks a render
- Pointing the gun at the camera. A pistol aimed toward the lens gets the clip stopped by safety review far more often. Aim it side-on, across the gap.
- Two people of the same gender. The model swaps faces unless you describe each person: "the woman with the short blonde bob and denim jacket", "the woman with long black braids and a red hoodie".
- A light zombie description. "Grey skin, white eyes" comes back as an ordinary person. Spell the look out and repeat it in the beats.
- Real names. Never name a real person or celebrity in the prompt; the model refuses public figures.
- Forgetting "no dialogue". Veo adds speech if you don't rule it out.
Or skip the prompts
If you'd rather not write, render twice and edit, the zombie AI video generator does all of this from two photos: it describes both of you, renders both reels and splices them for you. Your first film is free.
