AI Zombie Pet Video
An AI zombie pet video from two photos: the four shots, what holds a dog or cat’s identity, the reference pair that works, and why the animal reaches its owner without ever attacking.
Last updated: 2026-10-13
A dog stands frozen at the far end of a dark cabin, then runs the last few metres straight into its owner's arms. The owner has let the gun fall out of their hand. The embrace hard-cuts into a sunlit wheat field with the dog healthy again. Fifteen seconds, vertical, scored.
That is the pet version of the AI zombie trend, and it is harder to render than the human version. Video models are trained on far more human faces than dog faces, so the thing that usually breaks is not the emotion — it is the breed. The animal that walks in is a recognizable version of your pet; the one standing in the wheat field at the end is a generic brown something.
An early pet render from the examples page, in the cabin world. The dog's stride and the leap are the film's own; the beat map below is the story the generator pins today, which is the same crossing at the same seconds. In neither version does the animal attack, bite, or touch anyone with its teeth.
The four shots, in order
The pet subject keeps the same skeleton as the person subject — same four beats, same seconds, same match cut — and the world you pick (cabin, grey shore, rainy street, snowy doorstep) is shared by both. Only the animal-specific lines change.
| Shot | Seconds | What happens |
|---|---|---|
| 1 — The cabin | 0–4 | The owner stands trembling, gun lowered at their side, unable to raise it. Cut wide: the pet waits at the far end of the room, motionless, facing them — not attacking. |
| 2 — The undead face | 4–6 | Extreme close-up. Pale grey fur, cloudy white eyes, dark veins, disheveled coat. Frightening, and still unmistakably your animal. |
| 3 — The choice | 6–10 | The owner lets the gun fall and opens both arms. Something familiar flickers in those cloudy eyes. The pet runs the last few metres and presses its head and body against their chest. |
| 4 — The match cut | 10–15 | Mid-embrace, hard cut: the cabin becomes a golden wheat field at sunset. Both are alive and healthy, the pet's breed, coat and size restored exactly as in the reference. |
Cold blue-grey light for the first ten seconds, warm gold for the last five. The colour change does most of the emotional work; the performance only has to stay still enough for the cut to read as one hug.
What actually holds a dog or cat’s identity
Four shots, two lighting setups and a hard cut is a lot of chances for an animal’s face to drift. The identity sentence has to name the things a model can latch onto:
- Breed, coat, size — stated once, early, and repeated inside the final shot, which is where drift shows up.
- Facial identity — muzzle shape, ear set, eye placement. “The same dog” is not an instruction; “the exact breed, coat, size and facial identity of the pet” is.
- The undead look is painted on top, not substituted. Grey fur and cloudy eyes are layered onto that animal, so shot 2 is still recognisable.
- The owner is locked too. Their face, hairstyle and body proportions hold against the first reference. A film where the dog is right and the human is wrong still looks broken.
If you write prompts yourself, the free prompt page walks through the same beat map line by line.
The two photographs that work
Everything downstream depends on these two files.
Reference one — you. Shoulders-up, face toward the camera, daylight or an even room light. No heavy filter, no sunglasses, no hat. The model has to re-render this face in four shots, so a clear front view does more than a flattering one.
Reference two — the dog or cat. A full-body photo taken at the animal's eye level works best, because one frame has to carry both halves of the identity: the face (muzzle shape, ear set, eye placement) and the body (breed, coat, size). One animal in the frame, and nothing covering the face — a bandana, a costume, a toy held in the mouth, or a paw over the nose all remove features the model has to keep. Standing over the pet and shooting down loses both the face and the proportions.
Two things help more than people expect: similar brightness in both photos, and a recent photo of the pet rather than a blurry one from years ago.
Where pet renders fail, and the line that fixes each one
| What you get | The fix |
|---|---|
| A different breed in the wheat field | Restate breed, coat and size inside the last shot, not only at the top |
| The pet rises onto two legs or shuffles like a person | Add “moves on all fours” to the approach line |
| The pet lunges, or the scene reads as an attack | Keep both guards: “but not attacking” on the approach, “does not attack” before the embrace |
| The owner’s face drifts | The owner needs their own lock — face, hairstyle, body proportions against the first reference |
| A second animal wanders into frame | Put exactly one animal in the reference photo and state “no other animals” |
| The ending looks like two unrelated clips | Put the continuity instruction on the embrace itself: same arms, same distance, same direction of movement |
Why the undead pet comes closer instead of attacking
Two reasons, and the second is practical.
The story is grief and recognition, not threat. An animal that charges turns the third shot into a chase, and then opening both arms means self-defence rather than a choice. The approach is what the hug is answering.
The other reason is that a film with no gore, no biting and no blood is what most video models and payment networks will actually process. The restraint is not timidity — it is what keeps the format publishable, and it is why the trend reads as a love story dressed as a horror clip rather than the reverse.
Questions people ask about AI zombie pet videos
Can the pet be a cat?
Yes. The pet story is written for a dog or a cat matching the second reference — the identity line names breed, coat, size and facial identity, which are the same four things whether the animal has ears forward or a shorter muzzle. Use a photo where the face is as clear as the body.
Can I use a photo of a pet that has died?
A single photograph is all the render needs, and plenty of people make this film about a pet they lost. Nothing about the output claims the animal is real again — it is fifteen seconds of light in a wheat field. If you are including yourself in the first reference, that is your own likeness to consent for.
Can both references be pets?
No. The survivor in the first shot is always a person — the story is built on someone lowering the gun. Two animals produce a film with nobody to make that choice.
How long is the video, and what shape?
Fifteen seconds, four shots, vertical 9:16 — the shape Reels, Shorts and TikTok want — with ambience and score rather than a silent clip.
Do I have to write a prompt?
No. Our generator has no prompt box: the four-shot beat map is already pinned, and the only decision is which two photographs go in. If you would rather drive a third-party model yourself, the prompts and the reasoning behind each line are free on the prompt page.
What does a render cost?
One film costs one 480p or 720p tier price from a one-time credit pack — no subscription, and a failed render refunds automatically. Current amounts are on the pricing page.
Or make one now
Open the generator, pick the pet story, drop in your photo and your dog or cat’s photo. That is the whole workflow — the beat map, the identity lock and the no-gore clause are already handled.
If you want to see a rendered pet film before paying for one, the examples page carries the frames and clips from this exact pipeline.