How to Make an AI Zombie Video From Two Photos
The three beats, the two photos that decide everything, and the six failures to avoid — a step-by-step guide to making an AI zombie video.
Last updated: 2026-10-09
An AI zombie video is three beats and two photos. You upload a picture of yourself and one of someone you love, and the clip that comes back plays the standoff, a cut to the two of you in warm daylight, and the gun coming down. Nothing here needs a camera or an editor — a phone browser is enough, and the whole thing takes about ten minutes of your attention.
This page is the process in order: which photos work, which scene to pick, what to check before you post, and the six failures that account for nearly every bad result.
If you'd rather see the finished shape first, the format breakdown walks through the three beats shot by shot.
Step 1 — Choose the Photos
The photos decide more of the result than any setting on the page. Get this step right and the default settings are fine.
What works:
- One person per photo. Just you in one, just them in the other.
- Facing the camera, eyes open and visible. Not looking away, not mid-blink, not in shadow.
- Similar framing in both. Headshot with headshot, or waist-up with waist-up.
- Even light. Daylight or a bright indoor room. Nothing artistic.
What fails every time:
- Group photos. The model picks a face you didn't intend, or blends two.
- Screenshots, or photos taken off a screen. Moiré and glare make the eyes unreadable.
- Heavy beauty filters. Smoothing and slimming change the face, and the clip comes back looking like someone like that person — which defeats the whole point.
- Sunglasses, hats, or hair across the eyes.
- One photo tight, one wide. This is the single most common cause of a clip that looks like a slideshow.
A phone photo from last summer is ideal input. You don't need a good photo — you need a clear one.
Step 2 — Choose a Scene
Three scenes cover almost everything:
- Dim interior, cold overhead light. The default. The model has seen this setup more than any other, so it generates most reliably.
- Wet city street at night. Strongest contrast against the daylight beat.
- Open field at dusk. Softest; the right choice for the pet version and for memorial clips.
Do not put violence in the description. Blood, wounds, gunfire, "attacking" — these words come back in the video, and they move the clip from sad to horror. Those are two different formats and this one only works as the first.
Step 3 — Generate, Check, Export
- Generate. Usually thirty seconds to two minutes.
- Check the faces first. If they aren't recognisable, change the photo — not the prompt. Nine times out of ten the prompt was never the problem.
- Check the order. Standoff → daylight memory → lowered gun. Out of order, re-generate.
- Check the daylight beat. Any grey skin, clouded eyes or torn clothing leaking into the sunlight half is the most common failure of this format, and it can't be fixed in an editor. Re-generate.
- Export vertical. Export 9:16 unless you chose 16:9 in the studio. Credit packs are on the pricing page. Don't post the version with the framing you haven't checked.
When It Goes Wrong
| What you see | What actually caused it | Fix |
|---|---|---|
| Doesn't look like the person | Photo problem, not a prompt problem | Swap for a front-facing, evenly lit photo |
| Looks like a slideshow | The two photos are framed too differently | Headshot with headshot, then re-generate |
| Grey skin in the daylight beat | The model carried the transformation across the cut | Re-generate; it will usually clear |
| The clip drags | The standoff shot is too long | Ask for a shorter first beat, or cut it after export |
| One person's face is wrong in one shot | A second face in one of the photos | Crop the input down to one person |
| Nothing generates | File too large or a HEIC the browser can't read | Re-save as JPG |
What It Costs
There's no free tier and no subscription. You pay for the generations you actually run, which is why the queue moves at video-generation speed instead of being rationed behind a free allowance. One payment covers one video; length is 10 or 15 seconds at 720p, and packs are on the pricing page.
Output quality depends on the model and on your photos, and the photos are the part you control. That's why this whole page is about the input: read it before you pay, not after. Photos outside the list above are a coin flip.
FAQ
Is it free? No — there's no free tier, and no subscription to cancel. You pay per video; the pricing page has the current numbers.
What if I don't like the result? The output depends on the model and on the photos you upload. We don't offer refunds based on how much you like the video, and we don't promise a specific likeness or a specific look — that's why the photo rules and the sample gallery come first, before payment. If a file is broken, the payment goes through without a video, or the export fails, that's a different case: write to support@zombieai.video and we'll sort it out.
Do I need two photos, or does one work? Two. One of you, one of the other person (or your pet). A single photo can generate something, but the standoff beat needs both faces.
How long does it take? Generation runs thirty seconds to two minutes; the whole thing, including picking photos, about ten minutes.
Can I do this on my phone? Yes — upload and export both happen in the browser.
Can I use it commercially? Yes, provided the photos are yours to use. Don't use anyone else's footage or an unlicensed audio track; the terms cover it.
What if the person in the video has died? The clip is a generated scene, not a recording, and it won't move or speak the way they did. If that would make it harder rather than easier, don't make one. Plenty of people do make these — the pet page has notes on the version people use for someone they've lost.
Two photos, one scene, about ten minutes. Start a generation.