AI Zombie Love Story: How the Two-Photo Trend Works | CircleAIDance

The zombie love story is the AI video trend that turns two ordinary photos into a short, cinematic scene: one person aims at the other, who has turned; the shot never comes; the two embrace — and a hard cut lands the pair in a warm, ordinary memory. It spread across TikTok and Instagram through late September and early October 2026, and it is built to be remade: the beat sheet is short, the emotional turn is the whole point, and the edit that sells it is a single cut. This page breaks down how it works and gives you prompts you can paste — with the honest version of what our tools do and do not render in one go.

The beats, in order

Nearly every version follows the same four beats, and the order is the trick:

1. The standoff. A survivor in a ruined place raises a weapon at someone they know — now changed, but still recognisable. Dust, low light, a held breath.
2. The refusal. The weapon comes down. The survivor opens their arms.
3. The embrace. The changed one closes the distance — and the moment of contact is where the cut hides.
4. The memory. A hard cut to the same two people, healthy and warm, in daylight. Same faces, same pose, different world.

Both faces must stay recognisable throughout — that is what makes it land — which is why the photos you start from matter more than the prompt.

Why one hard cut carries the whole thing

The trend works because the cut does the arguing: the embrace in the ruined world and the embrace in the memory are matched frame-to-frame, so your eye reads them as one continuous moment rather than a scene change. Everything before the cut is setup; everything after it is the payoff. When you plan the clip, plan the cut first: decide the frame where contact happens, and make both shots arrive at that frame from the same angle.

How to make one from two photos

Here is the route that matches what our tools actually render. Two photos go in; the finished piece is two short clips joined at one cut. Nothing to install — it all runs in the browser.

1. Turn two photos into one scene. Our image-to-video animates a single reference image, so the first job is to compose both people into one frame. Open image-to-image: add both photos as reference images (Nano Banana 2.1 accepts up to ten), and prompt a single cinematic still of the standoff — same two people, one ruined interior, one frame. Keep both faces clear and uncovered; head-and-shoulders or waist-up photos with even light work best.
2. Animate the standoff. Send that still to image-to-video: one reference image in, the motion written in the prompt, and Veo 3.1 renders 6 or 8 seconds at 720p or 1080p with synchronized audio — the lowered weapon, the charge, the embrace, with the sound arriving in the same pass. Shoot it vertical (9:16) if it is headed for social.
3. Make the memory shot. Compose a second still — the same pair, healthy, in warm daylight, photographable happiness — and animate a short, gentle beat: a laugh, a turn toward each other, a slow drift.
4. Join at the embrace. Put the two clips in any editor, find the contact frame in each, and cut hard from one to the other. That cut is the trend; everything else is taste.

Two ground rules before you start. Use photos you have the right to use — both people should be fine with appearing, and if a photo is not yours, get the owner's permission; that is also what our terms require for anything you submit. And keep it at the cinematic level: our terms do not allow graphic violence or gore, and the trend loses nothing without them — the "no blood, no gore" line belongs in your prompt, below.

Prompts you can paste

Four sets, each with a still prompt for image-to-image and a motion prompt for image-to-video. Swap in your own details — clothing, room, weather.

1 · The standoff still (image-to-image, both photos as references)

One cinematic film still, vertical 9:16. Two people in the same frame inside an abandoned, dust-filled room: [person A] stands in the foreground holding a flashlight raised defensively toward [person B], who stands several steps away — face pale and marked, clothes torn and dusty, but clearly the same person, fully recognisable, eyes soft rather than menacing. Cold grey light through a broken window, floating dust, shallow depth of field. Sombre, restrained, like a drama about grief, not a horror film. No blood, no gore, no wounds. No text, no watermark.

2 · The lowering and the embrace (image-to-video, the standoff still as the reference image)

Begin on the standoff. [person A] slowly lowers the flashlight and opens both arms; [person B] rushes forward and collides into an embrace; both hold tight as the camera pushes in slightly and handheld shake softens. Keep both faces recognisable and facing camera at the end. Cold light, floating dust, restrained and emotional, no blood, no gore. Sound: quiet room tone, footsteps, fabric, one shaky exhale.

3 · The memory still (image-to-image, both photos as references)

One warm cinematic film still, vertical 9:16, the same two people in the same pose as an embrace — but healthy and happy in golden-hour daylight on an ordinary street. [person A] laughing mid-motion, [person B] mid-laugh and turning toward them. Soft backlight, warm tones, gentle lens bloom. Feels like a photograph someone kept. No decay, no marks, no text, no watermark.

4 · The memory beat (image-to-video, the memory still as the reference image)

A gentle handheld moment: the two hold their embrace in warm golden-hour light; [person B] turns slightly toward camera and laughs; [person A] closes their eyes. Camera drifts slowly sideways, film-like grain, shallow depth of field. Sound: street ambience, distant birds, soft laughter.

Variants that travel well. Swap the second person for a pet and the same beat sheet works — the trend's cat and dog versions do exactly this. For the no-decay version some people prefer, drop the pale-and-marked line from the still prompt entirely; the cut still carries it.

What a run costs

Every run bills, and the composer shows the exact credit estimate before you start — there is no free tier and no subscription. On Veo 3.1, the clips above render at 30 credits each for Lite at 720p (35 at 1080p), or 60 for the Fast tier at 720p (65 at 1080p); 6-second and 8-second clips cost the same. The two stills are billed per image at the image models' rates. A complete two-shot story is two video runs plus two stills — full rates are on the Veo 3.1 model page and pricing.

FAQ

Do I need two different photos?

Yes — the format needs two people, and usually two separate photos so each face reads clearly. Photos of the two of you together in one frame also work as references; the point is having both faces clear and uncovered, lit evenly, in head-and-shoulders or waist-up framing.

Why is my version not 15 seconds long?

Because the dedicated tools built for this trend ship a fixed-length cut — commonly 15 to 20 seconds — while general image-to-video renders 6 or 8 seconds per run. Ours is the second kind: you get the standoff-and-embrace in one run and the memory in another, and you join them at the cut — which is the part of the trend that matters anyway.

Can I make one of someone else?

Only with the right to use their photo, which includes them being okay with it. Our terms require that you hold the necessary rights to everything you submit, and likeness rules vary by where you and they live. When in doubt, ask.

Is the gore allowed?

No — our terms prohibit graphic violence and gore, and the trend does not need it. The version that performs is the restrained one: dust, cold light, and the turn. Keep the "no blood, no gore" line in every still prompt and motion prompt.

Do I need to install anything?

No. Both steps — composing the stills and animating them — run in the browser in CircleAIDance Studio; the only thing you bring is the two photos and the prompts above.

Keep reading

• PhotoCraft — the free open-source image editor from the ArtCraft team, and the browser AI route for photo edits.
• EffectCraft — the same team's motion graphics app: how to install it, a first-project walkthrough, and the timeline-free AI route.