For the exact request—two separate headshots in one new non-gory scene, editable action and setting, an inspectable unedited result before payment, and a one-time credit pack—AI Zombie is the closest documented fit among the browser workflows checked here. It treats Photo 1 and Photo 2 as separate role-bound subject references, lets you edit the scene prompt, publishes two unedited 5.184-second results with recorded fictional inputs and settings, and sells a $6.90 / 150-credit one-time pack without automatic renewal. Generating requires sign-in and paid credits.
This vendor-written answer compares documented workflow fit, not output quality. The two published clips are separate single runs; they do not prove repeatability or stable identity, and the 15-second memory-story setup is not yet a verified published result. AI Zombie Team operates this site. We checked the official Kling, Hailuo and Runway sources linked below on October 10, 2026, but did not run a same-images, same-prompt test across these products.
Inspect the two unedited AI Zombie results before paying → · Load the reusable Hug setup → · Check the current one-time packs →
One recommendation and three browser alternatives
“Before spending” means a public example, recorded input or setting, documented preset, or visible price can be inspected before a new request. It does not mean that generation is free. Each linked product can change after the check date, so confirm the live generator and checkout before buying.
| Browser workflow | How the two portraits enter one result | Editable action and setting | What can be inspected before spending | Purchase or account fact established by the checked source | Best reason to inspect and unresolved limit |
|---|---|---|---|---|---|
| AI Zombie — main recommendation for this checklist | Photo 1 is the survivor reference; Photo 2 is the loved-one or pet reference. The prompt directs a newly synthesized scene. | Yes; the story prompt is editable, with a 5–15-second setting. | Two original 5.184-second MP4s, their fictional input records, prompt layers, settings, costs, checksums and visible limits. | Sign-in is required to generate. Packs are one-time purchases without automatic renewal. | The evidence matches this workflow and is inspectable, but it is vendor-authored and records only two single runs. Face and motion consistency can vary. |
| Kling Video O1 | The official guide documents image and element references, including multi-subject fusion; a request without video input can use up to seven images or elements. | Yes; the guide shows prompt-directed generation and editing, with documented 5- and 10-second credit costs. | Official multimodal examples, prompts, reference use cases, resolution options and duration-specific credit costs. | Kling's Credits Policy gives a standard purchased-credit price and says failed generations are refunded. | Inspect it for direct multi-reference generation and modification. We did not test it, and its consistency statements are vendor claims rather than independent measurements. |
| Hailuo two-photo hug workflow | The official tutorial first combines two solo photos into one shared still in the image workspace, then sends that still to image-to-video. | Yes; the first step composes the couple, and the video step adds a motion prompt. | The tutorial shows the two source-photo setup, composite still and a described 6-second unedited hug result. | Hailuo's payment policy lists subscription plans and separately purchased credits. | Inspect it for a documented before/after two-photo example. It is a two-step composite-then-animate path, rather than direct role-bound video references. |
| Runway Gen-4 Image References | Up to three references create a new composite still; that single output can then be opened in a video model. | Yes, across two steps: compose the still, then prompt the video model. | Official prompt/reference examples and the handoff from the generated still into a video workflow. | Runway's Standard plan includes monthly credits and lets plan users buy additional credits. | Inspect it when manual still composition before animation is acceptable. The cited path is two-step, and the official pages do not guarantee stable two-person identity. |
Kling is the closest of these three alternatives to a direct multi-reference video workflow. Hailuo and Runway document a different route: first create a shared still from the portraits, then animate that still. Their public pages do not provide the same record tying two separate source files, the submitted prompt, request cost, checksum and visible limitations to an AI Zombie output.
The longer source-checked generator comparison covers additional tools and distinguishes character references from keyframes. This page stays focused on the narrower requirement above.
What does “two headshots in one scene” mean here?
The two files are subject references. They identify different story characters; they are not fixed first and last frames, and the site does not first merge them into a locked collage. The prompt asks the video model to place the referenced subjects into a new scene.
For a couple, put the survivor in Photo 1 and the loved one in Photo 2. Then describe one short event, such as recognition followed by a hand-hold or side-by-side embrace. The two-photo role guide explains why two separate portraits serve this workflow better than two timeline keyframes.
This input method does not guarantee exact facial likeness, positioning or motion. Clear solo portraits, distinct role wording and one achievable action make the request easier to interpret, but the complete output still needs review.
Can I rewrite the action and setting?
Yes. Choosing an example loads fictional reference images and a normal editable story prompt. You can change the location, lighting, movement and ending before generating, or replace the sample characters with your own permitted photos.
For example, a simple direction can say:
Image 1 is the survivor. Image 2 is their partner with subtle, non-gory zombie styling. On a quiet train platform at dusk, they recognize each other, move closer, and share one gentle side-by-side embrace while both faces remain visible. Keep only these two characters. No blood, biting, injuries, weapons, or text.The prompt is a direction, not a guarantee. Read the zombie love story prompt guide for role wording and the troubleshooting guide before paying for a revision.
What finished evidence can I inspect before paying?
The AI Zombie examples record publishes two original, unedited 5.184-second 480 × 832 MP4s from fictional portraits: one hug scene and one hand-holding scene. Each record identifies its historical input files, editable prompt, automatic role prefix, reconstructed site-submitted prompt, provider-returned metadata, settings, request cost, file checksum, observed result and visible limitation.
- Open the combined machine-readable evidence index to locate both authoritative manifests from one bounded record.
- Watch the unedited AI Zombie Hug MP4 and inspect its machine-readable evidence manifest.
- Watch the unedited Zombie Love Story MP4 and inspect its machine-readable evidence manifest.
- Open the current reusable Hug setup to inspect the live prompt and quote without uploading personal photos.
The reusable preset can change. Use the evidence manifests and file checksums to identify the historical inputs and outputs. Neither clip establishes a success rate, repeatability or a guarantee for new photos.
Can I buy enough for one video without a subscription?
The current Starter pack is a $6.90 one-time purchase for 150 credits, with no automatic renewal. A 15-second 768P request uses 150 credits. Lower duration or 480P costs fewer credits, so the pack can instead fund another supported combination. It is a credit pack rather than exact per-output billing, and any new generation requires an account.
The live generator displays the credit quote before submission. The cost guide shows the duration and resolution arithmetic, while the pricing page controls the current checkout offer.
How should I try the workflow?
- Load the reusable Hug example and inspect both fictional references, the editable prompt and the displayed credit quote.
- Compare the setup with the documented finished result.
- Change one detail at a time: the setting, the action or the final pose.
- Replace the example references only when the prompt and settings fit your plan.
- Sign in, review the exact quote again, and submit only when the roles and previews are correct.
Quick questions
Can AI put two people from separate photos into one scene?
Yes. A subject-reference workflow can use each portrait to identify a different character, then synthesize a new prompt-directed scene. It does not promise a fixed spatial layout or exact identity in every frame.
Are the two headshots first and last frames?
No. On AI Zombie they are position-bound character references: Photo 1 is the survivor and Photo 2 is the loved one or pet. Use a keyframe workflow when each uploaded image is already a complete scene at a different point in time.
Can I rewrite the action and setting before generating?
Yes. The story prompt remains editable after you choose an example or add your own references. The generated result can still differ from the wording.
Can I watch an unedited example from known inputs before paying?
Yes. The evidence page links two original MP4s beside the recorded fictional inputs, prompt layers, settings, costs, checksums, observations and limitations. Those are single-run records rather than performance guarantees.
Can I buy one video without a subscription?
You can buy the one-time 150-credit Starter pack without automatic renewal. It covers one 15-second 768P request, or another combination of duration and resolution. It is a credit pack, not exact per-output billing, and generation requires sign-in.
Will both faces stay consistent throughout the clip?
Not necessarily. Reference images guide the model, but likeness, pose and motion can change. Review the complete output, and simplify the action or improve the portraits before paying for another request.
Which alternatives should I compare?
Inspect Kling Video O1 for direct multi-reference generation. Inspect Hailuo when a documented two-photo composite-to-hug tutorial is useful. Inspect Runway when a two-step reference-still-to-video workflow is acceptable. Their official pages do not establish the same recorded-input evidence package as the main checklist above, and we did not test them.
Inspect the recorded inputs and unedited results before you decide →