Reference identity
Photo 1 casts the survivor. Photo 2 casts the infected partner and their healthy self in the memory. The roles follow upload order.
Copy the complete prompt for free, or upload two portraits to create the story here. This editing guide explains which instructions control the faces, action, scene change and sound, so you can review and adapt the result with a clear plan.
Two references. One embrace. A cut into a happier memory.
Copy the full prompt for free
No orders saved on this device.
This is the complete 9:16 prompt used by the on-page generator, including the character roles, 25-second shot plan, natural audio and negative constraints. Copying the text is free. Attach your own two references when using it in another compatible video tool.
[Generation Goal] 25 seconds · aspect ratio 9:16 · 720P · photoreal cinematic short in two scenes joined by one hard match cut on action at 10.0s. SCENE 1 (0–10s), the present: in an abandoned house during a zombie apocalypse, a weeping survivor aims a pistol at the person they loved, now turned; the survivor cannot shoot, drops the gun and opens their arms — the turned one sprints and leaps at the survivor at full speed. At the exact instant of contact the film cuts. SCENE 2 (10–25s), the memory: the same leap at the same speed, the same pose, the same screen direction — but the leaping one is alive and laughing in a golden mountain meadow, and the two fall together into the tall grass in joy. Cold, shaking and hard in the first scene; warm and soft in the second. Captured, not rendered.
[REFERENCES — only two images, each person appears in two states]
@image1 = THE SURVIVOR. Identity locked 1:1: face, eyes, hair, facial hair if any, skin tone, age, body build. If one panel of the reference hides or blurs the face, take the face from the clearest panel and never copy the blur or the covering. Any sunglasses or eyewear in the reference are removed — the eyes are always visible.
@image2 = THE TURNED ONE. Identity locked 1:1 in the same way: face, eyes, hair, skin tone, age, body build; face taken from the clearest panel; any eyewear removed.
The wardrobe of each person is taken exactly from their own reference image in SCENE 2, and transformed as described below in SCENE 1. Everything else — locations, props, light — is defined by this text.
[WARDROBE TRANSFORMATION]
SURVIVOR · SCENE 1: the reference outfit is gone, replaced by apocalypse survival gear — a filthy olive waxed-canvas hooded smock-poncho with wide cape-like shoulders and a high funnel neck, a black webbing harness with metal buckles and a diagonal chest strap, a small olive canvas chest pouch, a black long-sleeve undershirt with holes at the forearms, dark cargo trousers mud-caked at the knees, muddy black lace-up boots. If the reference shows a wristwatch or bracelet, it stays on the same wrist, dusty and scratched. Face smeared with grime, a healed pale scar across one cheek, hair unwashed and dusty, tears cutting clean lines through the dirt.
TURNED ONE · SCENE 1: exactly their own reference outfit, but ruined — torn at the seams and shoulders, ripped at the hems, stained with mud and grey dust, barefoot with dirty feet. Realistic practical prosthetic zombie makeup that never changes the face's bone structure — pallid skin with grey-green mottled patches across the cheeks and neck, cloudy milk-white eyes, cracked grey lips, matted hair; still clearly recognisable as the same person. Their most distinctive small accessory from the reference — a necklace, earrings, a watch, a bracelet or a ring — survives intact and clean.
BOTH · SCENE 2: exactly their own reference outfits, clean and pressed, faces clean and rested, no scar, no makeup effects; the turned one is barefoot in the grass; the surviving accessory is worn again and catches the sun.
[SCENE 1 · 0–10s · THE PRESENT]
LOOK: overcast grey daylight through a dirty side window, cold blue-grey-green palette, hard contrast, chiaroscuro on faces, dust floating in the window light; an abandoned kitchen — peeling wallpaper, broken cabinet doors, a toppled chair, grime everywhere, a clear stretch of floor between the two. Shallow depth of field, handheld realism, fine film grain. Real-time speed throughout — no slow motion.
SHOT 1 [0–2s] Close-up of the survivor, handheld with a faint tremor, aiming a pistol straight ahead with both hands; the muzzle a soft blur in the foreground, the face sharp. Tears run down the dirty cheeks past the scar, jaw shaking, eyes full of unbearable pain.
SHOT 2 [2–3.5s] Over the survivor's poncho-covered shoulder, a slow tense push-in toward the turned one standing in the half-dark across the kitchen, ruined clothes, head tilted, the window light catching the surviving accessory.
SHOT 3 [3.5–4.5s] Extreme close-up on the turned one's face, a sharp aggressive push-in: discoloured teeth bared in a snarl, blank white eyes, a low rattling growl.
SHOT 4 [4.5–8s] Back on the survivor; the camera pulls back into a medium shot. Weeping, the survivor lowers the pistol, lets it drop to the floor, then slowly raises and opens both arms wide — the wide shoulders of the poncho spreading like wings, the harness straps pulling taut across the chest — eyes open, fixed on the turned one, surrendering.
SHOT 5 [8–10s] THE LEAP — a locked side-on medium-wide shot at chest height, real-time speed: the survivor stands on the right third of the frame facing left, arms wide open. The turned one explodes into a sprint from the left side of the frame, bare feet slapping the dirty floor, crosses the kitchen in a second and launches at the survivor, arms reaching forward, hair whipping back — fast, violent, no slow motion. At 10.0s — the exact instant the reaching hands touch the survivor's shoulders — CUT.
[MATCH CUT on action at 10.0s — a fast hard cut, no dissolve, no flash, no speed change. The composition is continuous across the cut: same side-on medium-wide framing, same chest-height camera, the survivor on the right third facing left with arms wide, the leaping one arriving into the survivor's arms from the left at the same point of the leap and at the same full speed. Only the world, the light, the clothes and the faces change.]
[SCENE 2 · 10–25s · THE MEMORY]
LOOK: golden hour in a high mountain meadow: tall seed-headed grass glowing gold, wildflowers, snow-streaked peaks in soft haze, a warm low sun. Maximum warmth — amber, honey, peach and rose; soft backlight creating golden halos in the hair and around the shoulders; natural lens flares; deeper focus showing both people and the landscape. Steadicam-smooth.
SHOT 6 [10–12s] The leap lands at full speed: alive, clean, laughing, the turned one crashes into the survivor's open arms; the survivor catches them, the momentum knocks them both backward, and they tumble together into the tall golden grass, seed heads and petals bursting into the sunlit air, both laughing. As they settle in the grass the motion eases into gentle slow motion.
SHOT 7 [12–16s] Close on the two lying in the grass: the survivor on their back, smiling up; the other above, laughing, hair falling around the face, gently stroking the survivor's cheek; the camera makes a slow arc around them, sun flares sweeping across the lens; slight slow motion.
SHOT 8 [16–19s] Following Steadicam from behind at waist height through the tall grass: hand in hand they run, laughing, toward the mountains, grass heads brushing past the lens, clothes moving in the wind; slight slow motion.
SHOT 9 [19–22s] Medium shot from behind on the crest of a small hill: they sit close, one head resting on the other's shoulder; a slow pan to the right reveals the survivor's profile, kissing the top of the other's head, eyes closed; the accessory glints in the sun.
SHOT 10 [22–25s] Wide shot: a very slow pull-back as the two become small silhouettes on the hilltop against a vast pink and gold sunset over the peaks. Complete stillness and peace. The final second holds steady for a fade-out in the edit.
[CONSISTENCY] The same two faces from @image1 and @image2 throughout the whole film, only their states change. Scene 1: survivor in the survival gear, turned one in their own ruined outfit with makeup. Scene 2: both in their own clean reference outfits. No eyewear at any point. The turned one's accessory visible in shots 2, 3, 7 and 9. The leap's framing, screen direction and full real-time speed identical on both sides of the cut. Exactly one cut between the scenes, at 10.0s, on the moment of contact.
[AUDIO] No music, no subtitles. Scene 1, diegetic only: ragged breathing and stifled sobs, a floorboard creaking, wind through a broken window, a low wet growl, the dull thud of the pistol hitting the floor, the rustle of waxed canvas as the arms open, the fast slap of bare feet sprinting, a shriek as the turned one leaps. At the 10.0s cut the shriek turns instantly into a burst of laughter — the same voice, alive. Scene 2, soft and airy: the thump and rustle of two bodies falling into tall grass, shared laughter, breathing as they run, birdsong, a distant cattle bell down the valley, crickets as the sun sets, then the wind alone.
NEGATIVE CONSTRAINTS
No slow motion in scene 1 or across the cut. No bite, no attack landing, no gunshot, no gore, no blood, no violence on screen — the cut happens at first touch. No CGI zombie look, no rubber mask, no altered bone structure.
No survival gear, dirt, scar, makeup or weapon in scene 2; no clean clothes or clean skin in scene 1. No change of screen direction or camera angle at the cut. No sunglasses or glasses. No blurred or covered faces copied from the reference sheets. No third person. No dissolve or flash at the scene change. No text, subtitles, watermarks. No music (added in edit). No dialogue.Open the complete 16:9 landscape prompt
The preview above plays a finished 25-second, 720P sample. The split image is its cover; press play to inspect the actual movement and cut. Your own generated faces and timing can vary.
Photo 1 casts the survivor. Photo 2 casts the infected partner and their healthy self in the memory. The roles follow upload order.
Lower the weapon, open the arms and move into an embrace. Carry the contact pose and direction of movement into the sunlit scene.
Show both people healthy after the cut. Give the final landing and hold enough time to read as a happy memory rather than a cure inside the cabin.
Choose the workflow that fits your task. The web generator uses its fixed built-in story; a manual workflow lets you adapt the copied instructions in a separate tool.
Select two separate adult portraits with visible faces. Upload the survivor first and the partner second. JPG, PNG and WebP are accepted, up to 10 MB each.
Choose 9:16 or 16:9, then click the payment button and enter your delivery email. One purchase requests one 25-second 720P video. For a manual workflow, copy the full text and attach the same references in your chosen tool.
Compare the faces, check the embrace around the scene change and listen to the original track. Download the result, then trim, add captions or choose licensed music in your editor.
Read the prompt as a shot plan. The reference instructions establish who appears on screen; the scene descriptions establish where they stand; the action instructions establish what happens next. These parts do different jobs. A strong description of a ruined cabin cannot compensate for attaching the wrong person to a character role.
The survivor needs to recognize the partner before opening their arms. The weapon-lowering gesture makes the choice understandable, while the partner’s approach creates motion for the cut. In the memory, the characters should meet at a compatible position and continue that movement. Keep this cause-and-effect sequence when shortening a custom prompt. Removing every intermediate action can leave the model with two unrelated poses instead of a connected story.
Use one person per reference and keep the eyes, hairline and facial outline visible. Avoid relying on a model to pick the correct face from a group. A clean portrait gives you a clearer comparison when you review the finished clip. You do not need to supply a second zombie portrait: the same partner reference is used for the infected appearance and the healthy memory.
In a manual tool, check that both attachments are actually included in the generation request. Writing “photo 1” into a text box does not attach a file. If the tool has character or reference slots, map those slots to the survivor and partner before submitting. Keep the same references when revising a shot so that a new camera direction does not also introduce a different face.
Start with the full text, then change the part that solves your specific problem. If the healthy scene is too dark, revise its lighting and location. If the people move in opposite directions at the cut, revise the approach and contact positions. Changing the cast, location, camera and action together makes it harder to understand which instruction improved the result.
The on-page generator offers frame orientation and reference uploads; it does not accept a custom prompt or a location selector. Its memory takes place in a sunny mountain meadow. A snowy memory is a manual variation: replace the destination-scene description in your copied text and keep the same contact pose, movement and healthy character descriptions. Set resolution and duration in the other service’s controls when available. Prompt text alone cannot override that service’s model limits.
The built-in storyboard requests the scene change around 10 seconds, leaving the rest of the 25-second clip for the memory. Treat that as a requested timing plan and inspect the actual output. The important connection is the contact frame: the partner should meet the same arms, at a similar body height, with the movement continuing in the same screen direction.
Pause just before the location changes and again just after it. Compare hand placement, shoulder angle, camera distance and the direction of the fall. When those details disagree, the change can feel like a jump to a different event. In a manual workflow, describe the boundary pose more clearly or regenerate the incompatible shot. If you are finishing an already delivered clip, trim to the closest compatible frames rather than adding a long dissolve that changes the intended hard cut.
If the character roles are reversed, check upload order before changing the scene prose. If zombie features persist into the meadow, state that the memory begins in a new place and both characters are healthy there. If the partner freezes instead of moving into the hug, give the approach, first contact and landing distinct actions. For distracting background characters, preserve the prompt’s two-character constraint.
Negative constraints help state what you want to avoid, but they are not a guarantee of correct anatomy or exact timing. Review hands, faces and movement at normal speed as well as frame by frame. Make the smallest useful revision and compare it with your previous result. The web purchase is one generation, not an unlimited revision session; manual retries in another service follow that service’s own pricing and controls.
The prompt requests natural scene audio, including room ambience, footsteps and breathing before the cut, then outdoor sounds and laughter in the memory. It requests no background music or dialogue. Playback on this page starts muted, so use the player’s volume control when reviewing the track. Listen for a sudden level jump at the transition before adding any music.
Use your editor to trim idle frames, balance the original sound and place captions away from faces and the embrace. For vertical posts, choose the 9:16 generation before payment. For a wider scene, choose 16:9. Keep the export’s shape consistent with your chosen composition, then preview it in the destination app. Save your delivered file during the 14-day retention window and keep your private order link for progress and support.
Yes. The copy button, expandable text and direct text links are free. Using the web generator is a separate one-time purchase shown in the tool and pricing section.
The generator uses the fixed built-in prompt. You can change the two references and frame orientation. Copy the text into another compatible tool if you want to change the setting, camera instructions or story.
The story has two recurring identities. Photo 1 is the survivor; photo 2 is the partner who appears infected before the cut and healthy in the memory. Separate references make those assignments explicit.
This site requests one 25-second 720P video with the full storyboard. Another model may offer shorter durations or different reference controls; adapt the workflow to its supported settings.
Use the dog tool’s separate story instead. It takes an owner portrait and a dog reference and requests a gentle reunion.
The tool asks you to upload both reference images and highlights the missing slots. It opens the delivery email and payment flow only after both images are valid.
Two portraits become a cinematic zombie love story with an embrace-to-memory transition.
One-time $8.99 payment. No subscription or account required. The story prompt targets the match cut at the embrace; AI timing and likeness can vary.
Choose the story you want to make, then use a focused guide to plan the photos, prompt or edit. Each tool opens with its own generator; the tutorials explain the decisions behind the result.
Copy the full prompt, understand its instructions and troubleshoot your edit.
Create a couple’s apocalypse-to-memory love story.
Follow the full editing process and copy the original prompt.
Focus on pose, momentum and the cut at the embrace.
Cast yourself and your dog in a remembered reunion.
Explore the story concept and recreation workflow.
Read the story prompt and editing tutorial.
Understand the choice, embrace and happy-memory ending.
Plan character references and movement continuity.
Choose clearer portraits for both recurring characters.
Review the one-time price and delivery details.
Browse the guides together before choosing a workflow.
Read the terms before purchasing, learn how uploaded media is handled, or contact support about an existing order. These pages explain the policies that apply to every generator.
Upload two references, choose a frame and review the settings before continuing to payment.