Choose two photos for your zombie couple video
A practical guide to photo clarity, role assignment, supported preview files and what happens to your images.
Start with a clear picture of each person
The casting panel has two photo positions. The first is the survivor; the second is the zombie. Choose a separate picture for each person so that the assignment is unambiguous. You can use Swap roles afterward. There is no need to decide the final casting before looking at the two images together.
For preparation, prefer a recent photo where the face is easy to recognize, the lighting is even and another person is not competing for attention. A simple portrait is a practical starting point. This is guidance for choosing useful inputs, not a tested claim that a particular pose produces a better AI result. The model and its final image requirements are still undecided.
Only use photographs that you have permission to use, with permission from both people who will appear. Selecting the confirmation checkbox is your acknowledgement of that permission. It does not obtain consent from the other person or verify who owns a photograph. If you are unsure whether someone wants to appear in this story, resolve that before preparing their image.
Check light, detail and obstructions
Open the candidate photo at a comfortable size before selecting it. Can you clearly distinguish the eyes, nose, mouth and outline of the face? Strong backlighting can make those features difficult to see even if the overall image looks dramatic. A softer, evenly lit image usually makes the preview easier to assess.
Avoid selecting a tiny face from a very wide group photograph. Enlarging that area can make it look bigger without adding missing detail. If you have a more direct portrait, use that instead. The preview checks the dimensions of the full file, not the amount of useful detail in the person’s face, so passing its size check is not a quality assessment.
Sunglasses, hands across the face, heavy motion blur and strong beauty filters can change or hide identifying features. A photograph without those obstructions is easier to inspect. You do not need to remove ordinary personal details or make everyone look the same. The purpose is to choose recognizable people, rather than create a standardized passport-photo appearance.
Use files that the browser preview can open
The current preview accepts JPG, PNG and WebP images, with both dimensions at least 256 pixels. It does not impose a fixed file-size limit: the browser needs to be able to open your photo. The final service will adapt images to the selected model’s requirements before generation. Model-specific limits and recommended resolution have not yet been set.
If a phone photo is saved as HEIC, export a JPG copy before selecting it. Renaming the file extension does not convert its contents. Choose an export or save-as option in your photo application so that the image is actually encoded in a supported format. Keep the original photograph if you want to preserve the higher-quality source.
An image can have a supported extension and still fail to open. For example, the file may be incomplete or contain data that the browser cannot decode. The preview reports the problem and leaves the previous valid selection in place. Try another export of the photo, or choose a different picture; an error does not mean anything has been uploaded.
Assign the survivor and the zombie
The survivor is the person who faces the decision and opens their arms. The zombie is the person who approaches them. The reference story later returns to happier memories, where the two people are shown as a couple. Choose the assignment that feels right for the story you want to tell.
Roles are independent of gender. The panel does not infer a role from a face, a name or any personal characteristic. It simply uses the position you choose. Swap roles exchanges the two photo selections. If only one position contains a photo, the same control moves that selection to the other role.
Removing a photo clears that position in the preview. Selecting another valid file replaces the existing preview. Neither operation changes the source file on your computer. The small thumbnail uses a cropped display area, so the picture may appear trimmed on screen; it is not evidence that the final video will crop the person in the same way.
Know what the preview actually does
The preview opens your selected images locally and displays them in the casting panel. It does not send their bytes to our server, start a background generation task or charge credits. The generation control remains disabled because the complete video workflow has not yet been validated.
Your selections are held in browser memory for this page. Reloading the page, closing it or leaving the casting view clears the preview. There is currently no saved photo draft attached to an account, and choosing two images does not create a video entry in My videos. If you return later, choose the files again.
Selecting both images and confirming permission makes the cast ready to inspect. It does not guarantee the quality of a future render. The site does not currently measure facial similarity, predict identity consistency or rate your photo with an invented score. You remain responsible for checking that each selected image depicts the intended person.
Prepare for a fixed story rather than a custom prompt
This project is designed to preserve the chosen reference’s camera movement, action and edit while replacing both characters. You do not need a written description of a zombie, a battlefield or a happy memory. The intended personalization is the cast, not a new plot that happens to share the same theme.
The planned replacement includes appearance, hair, clothing and body. How well those details remain consistent across the full video still needs to be tested with the selected model. Do not spend time dressing or posing someone to satisfy an unverified requirement. Watch the reference and prepare clear images first; any further requirements will need to be published before paid submission.
The original reference music is the planned default. There is no need to attach an audio file to your photos. The examples on the homepage already include their source audio, so you can hear the timing while deciding whether the story fits. The eventual personalized soundtrack will be checked separately before the service opens.
Understand the future retention policy
When uploads become available, the intended retention period for an uploaded photo is seven days from its successful original upload. Derived copies, such as resized images prepared for processing, share that original expiry. Creating another copy should not restart the retention clock.
Finished videos follow a different rule: permanent retention, with no automatic age-based expiry, unless the owner deletes them. A short-lived download link can expire while the underlying video remains stored. Those two concepts should not be confused. The current preview has neither uploaded photos nor finished videos in cloud storage.
Before submitting a paid task in a future release, check the published price, the role assignment and the storage notice. No credit amount or turnaround time has been finalized for this preview. If the page still says creation is coming soon, you can inspect your selected photos, play the references and return later without making a purchase.
Watch the references
- CHOICE by @techmaf · Published 2026-09-16 (UTC)
- Another kind of forever by @damienstraker · Published 2026-09-26 (UTC)
Play both films on the homepage, or try the two-photo preview.