The zombie AI trend is a short two-person video: one of you has turned into a zombie, the other raises a gun and can't fire, and the clip cuts to a warm memory of the two of you before the gun comes down. You need two photos, one per person. The written tutorials make it in two generations and stitch them in CapCut; it can also be made as one 12-second take, with nothing to edit.
What it is: a Reels / TikTok / Shorts clip in four beats: the turn, the gun, the memory, the gun lowered.
What you need: two clear, front-facing photos. Photo one is the person holding the gun; photo two is the one who turns.
The usual way: two prompts in Flow or Gemini (zombie scene, then memory scene), joined in CapCut or VN.
The one-take way: all four beats in one prompt, one 9:16 clip, about two to three minutes on Kavel's zombie AI video page.
Copy a zombie AI prompt: five ready prompts are below. Swap who turns by swapping the photo order.
The zombie AI video page on Kavel, screenshot 4 October 2026.
What is the zombie AI trend?
The version going around in early October 2026 runs like this:
| Beat | What happens on screen | Rough timing in a 12 s clip |
|---|---|---|
| 1. The turn | One person has become a zombie: grey cracked skin, clouded white eyes, torn clothes | 0–3 s |
| 2. The gun | The other person raises a gun, shaking and crying, and can't pull the trigger | 3–6 s |
| 3. The memory | Cut to the two of them laughing together on a bright day, both looking like themselves | 6–10 s |
| 4. The gun lowered | Back in the dark room, the gun comes down | 10–12 s |
The four zombie AI beats, timed for a 12-second clip.
The couple is the usual cast, but not the only one: the example clips on the zombie AI video page include a husky as the one who turns.
Looking for the zombie filter instead? That one turns a single photo into a zombie face. The AI zombie filter does it from one photo. The trend video needs two people, because the story is about the one who can't shoot.
The common way to make the zombie AI trend: two prompts and CapCut
The written tutorial on PromptSeen (published 3 October 2026) splits the story into two generations and joins them afterwards. Here is that route, step by step.
- Pick two photos. One clear photo of each person, faces fully visible. You will use the same two photos, in the same order, for both parts.
- Generate part one, the zombie scene. Upload both photos to an AI video tool that takes reference images (the tutorial names Flow and Gemini). The prompt sets the scene, turns the second person into a zombie, has the first person aim a handgun, lower it and open their arms, and ends on the two of them falling backwards in an embrace. Note the order: in this version the gun comes down before the memory, not after it.
- Generate part two, the memory. Upload the same two photos again. This prompt starts from that mid-air embrace, changes the room into a golden-hour snow slope, and turns the zombie back into a healthy person for the memory scene.
- Join the clips. Put both parts in CapCut or VN, line up the embrace, and smooth the cut.
Where it goes wrong, in the tutorial's own words and in practice:
- Faces drift between the two clips. Each generation reads the photos again on its own, so the person in part two can come out slightly different from part one. That is why the tutorial insists on the same photos in the same order.
- The join has to match. Part two only feels like one story if it opens on the exact pose part one ended on. If your tool can't start from a frame, you line it up by hand in the editor.
- It is two runs and an edit. Two generations to pay for or wait on, and a third app for the cut.
If you already have Flow or Gemini credits and you like editing, this route works, and it gives you full control over the cut.
The zombie AI trend in one generation
The other way is to write all four zombie beats into one prompt and let one generation do the whole thing. That is how the zombie AI video page on Kavel is built.
The zombie AI video generator. Step 1: pick a template from the row at the top (each one fills in its prompt). Step 2: upload the two photos. The cost is shown above the button before anything runs. Screenshot 4 October 2026.
What changes compared with the two-prompt route:
- One clip, no edit. The turn, the gun, the memory and the lowered gun come back as a single video, so there is no join to line up.
- Upload the gun-holder first and the one who turns second. The second person then appears as themselves in the memory. The prompt names which photo plays which role.
- Vertical by default. The page opens on 9:16, the shape Reels, TikTok and Shorts want, so the clip is ready to post.
- Price before you press. The page runs on Seedance 2.0 Mini, the light member of the Seedance 2.0 family. A 12-second clip at 720p costs 925 credits; at 480p it is 430. One zombie AI clip at the standard settings can be bought on its own for $9.99, no subscription (plans are on the pricing page), and a clip that fails is refunded. A run takes about two to three minutes.
A one-generation clip can't be cut or retimed between the zombie scene and the memory; you steer it with the prompt, not with an editor's timeline. If you want a long memory sequence, go up to 15 seconds rather than adding a second clip.
Single-person zombie transformations are posted vertical too: this OpenClips community template is 9:16, 20 seconds, Seedance 2.5. Screenshot 4 October 2026.
Zombie AI trend prompts you can copy
Each zombie prompt below is written for two reference photos, uploaded in order. Photo one holds the gun; photo two turns. Paste one into the generator, upload your photos, and run it. To flip the roles, swap the upload order, not the prompt.
1. The abandoned house (the classic zombie AI prompt)
Two reference photos, one person each. Keep each face, hairstyle and build exactly as photographed and never swap them. Shot 1: in a dim abandoned house during a zombie apocalypse, the person from the SECOND photo has turned into a zombie, grey cracked skin, clouded white eyes, torn clothes, and stumbles forward. Shot 2: the person from the FIRST photo raises a pistol with shaking hands, tears on their face, unable to fire. Shot 3: a warm flashback of the two of them laughing in a sunny field, both looking exactly like their photos. Shot 4: back in the dark room, the first person lowers the gun. Cinematic film look, soft piano score, no dialogue, no other people, no text.
2. Snowed-in cabin
Two reference photos, one person each; keep both faces exactly as photographed. Inside a cold, snow-covered mountain cabin, the person from the SECOND photo has turned: pale grey skin, milky eyes, a torn winter coat. The person from the FIRST photo stands by the door holding a rifle, crying, and cannot shoot. Cut to a golden-hour memory of the two of them falling into deep snow and laughing. Back in the cabin, the rifle drops to the floor. Cold blue light inside, warm sunset light in the memory, 35mm film look, no dialogue, no text.
3. Gas station at night
Two reference photos, one person each; never swap the faces. At an empty gas station at night, flickering lights, the person from the SECOND photo walks out from behind the pumps as a zombie, clouded eyes and grey skin. The person from the FIRST photo raises a flare gun, hands shaking, and freezes. Cut to a summer road trip: the two of them singing in a car with the windows down, both exactly like their photos. Back at the station, the flare gun lowers. Neon and sodium light, handheld camera, quiet synth score, no text.
Whichever zombie AI prompt you paste, the page sends both photos as references, one per person. Screenshot 4 October 2026.
4. Hospital corridor
Two reference photos, one person each; keep identities exact. In a dark hospital corridor with emergency lights, the person from the SECOND photo, in a hospital gown, has turned: grey veins, white eyes. The person from the FIRST photo holds a pistol at arm's length, sobbing, and can't fire. Cut to the day they met: the two of them sharing an umbrella in the rain, laughing. Back in the corridor, the first person lowers the pistol and reaches out a hand. Cinematic, shallow depth of field, slow strings, no dialogue, no text.
5. Parent and child (no romance)
Two reference photos: the FIRST is the parent, the SECOND is their adult son or daughter; keep both faces exactly. In a ruined family kitchen, the SECOND person has turned into a zombie, grey skin and clouded eyes. The parent raises a pistol and can't fire. Cut to an old home-video memory: the two of them on a birthday, blowing out candles together, grainy and warm. Back in the kitchen, the parent lowers the gun. Film grain, soft piano, no dialogue, no text.
Why did my zombie AI trend video get blocked?
Under one of the zombie-trend tutorials on YouTube, a viewer wrote that their run was refused because the prompt said "pistol". Video generators screen prompts for violence, and this trend sits right next to the line.
- Show the gun, don't fire it. "Raises a pistol and can't fire" describes the trend. "Shoots", "bullet", "headshot" describe a shooting.
- Leave out blood, wounds and biting. Grey skin, clouded eyes and torn clothes read as zombie. Gore adds nothing the trend needs.
- Keep everyone an adult in the zombie scenes. Photos of children are a common reason a run with a dark theme is stopped.
- If it is still refused, change the object. A flare gun, a baseball bat, or just an outstretched hand keeps the beat: the person who can't bring themselves to do it.
On Kavel, a clip that fails is refunded in full.
Which way should you make the zombie AI trend?
| You want | Use |
|---|---|
| Full control over the cut, and you already have Flow or Gemini credits | Two prompts + CapCut |
| One finished vertical clip with no editing | One generation on Kavel's zombie AI video page |
| A longer memory section | One generation at 15 s, or the two-prompt route |
| A zombie face on a single photo, no story | The AI zombie filter |
| The same two-photo idea, different trend | The Hotel Lobby AI video |
Pick the route by what you want out of it.
FAQ
Is the zombie AI trend free?
Not on Kavel. The zombie AI clip is paid: 925 credits for 12 seconds at 720p, or $9.99 for one clip at the standard settings with no subscription. The two-prompt route costs whatever your video tool charges for two runs.
Can I do the zombie AI trend with one photo?
Not the trend itself, because the story needs a second person to hold the gun. With one photo, the AI zombie filter turns your face into a zombie.
Can I use a pet photo?
Yes: dog versions of the trend exist, with the dog as the one who turns. Put the pet photo second and say "dog" in the prompt where it says "person from the SECOND photo".
What app is the TikTok zombie trend made with?
There is no single official zombie effect. The clips are made in AI video apps from photos and a prompt; the PromptSeen guide uses Flow or Gemini for the clips and CapCut or VN to join them. You can also make it in a browser with no app at all.
Keep reading
- Make the zombie AI video from two photos
- Turn one photo into a zombie face
- Another two-photo trend: the Hotel Lobby AI video, and how the Hotel Lobby trend works
- Another tearjerker: the AI crying filter
- A zombie look for Halloween instead: the AI Halloween costume generator
- A vertical outfit trend: the ceiling fan transition
Sources
- PromptSeen, "Zombie to Love Story Memory AI Video Editing Prompt", published 3 October 2026, read 4 October 2026: promptseen.com/zombie-to-love-story-memory-ai-video-editing-prompt/
- OpenClips community, "Zombie Transformation — The Lunge", posted 30 September 2026, read 4 October 2026: openclips.ai/community/v/zombie-transformation-the-lunge-28828b07/
- Kavel zombie AI video page and generator pricing, read 4 October 2026.
