The Door to Anywhere: Walk Out of Your Room Into Any World
You film about three seconds of real footage. One screenshot from it becomes the start frame, one prompt builds the second half, and the two get stitched into a single shot. The whole thing runs inside ChatGPT.
Step 1: Film the door
Set your phone or camera on a tripod facing a closed door and leave it completely still. No handheld, no repositioning, no drifting. A locked off camera is what makes the illusion work, because the moment the frame moves the two halves stop lining up.
Walk into frame, reach for the handle, and open the door. Three or four seconds is plenty. What you actually need is a clean moment where your hand is on the handle and the door has just started to move.
If a cat wanders into your shot, leave it in. The imperfections are what make people believe the first half is real.
Step 2: Grab your start frame
Scrub back through the clip and screenshot the exact moment the door begins to open. That still is the start frame, and it is doing a lot of work: it carries your face, your clothes, the room, the lighting and the door position into the generated half.
Pick a frame where your face is clearly visible and nothing is motion blurred. A blurry start frame produces a second half where you look like a slightly different person.
Step 3: Open Higgsfield inside ChatGPT
In ChatGPT, open the plugins menu and select Higgsfield. Everything happens in the chat from here, and the model you want is Seedance 2.5.
⚡ Get the Higgsfield plugin for ChatGPT
Seedance 2.5 is the part that matters. It holds a real person's identity from a single still while building a completely different world around them, which is exactly what this effect needs.
Step 4: Paste the prompt
Upload your start frame, paste this, and send it. This version builds a Minecraft world on the other side of the door.
Use Seedance 2.5 in Higgsfield for this generation.
The first frame is exactly the attached photo, unchanged.
The person in the photo opens the door in front of them. Their face, hair, outfit, skin tone and body proportions stay exactly identical to the photo throughout, and the interior around them stays exactly as it appears in the photo. The camera is completely static and locked for the entire shot, no panning, no zooming, no camera shake, no movement of any kind.
As the door swings open, instead of revealing the interior background, the doorway opens directly onto a Minecraft world seen from human eye level: standing on a grass block at ground level, a vast survival world stretches ahead in bright midday. Everything is built from one meter cubes locked to a strict grid, grass blocks with green tops and brown dirt sides, stone, oak logs, cobblestone. Every surface uses low resolution 16x16 pixel textures with hard pixel edges, no smoothing, no anti-aliasing, no rounded corners, no bevels, no realistic material shading.
The world is spectacular. A deep ravine cuts across the middle distance with a waterfall of flowing water blocks pouring into it, and glowing orange lava spilling from an exposed cave mouth in the far wall. Beyond it, enormous floating islands hang in the sky with vines and thin waterfalls trailing off their undersides. Giant cubic oak and spruce trees rise from the valley floor, canopies made of blocky leaf clusters. On a plateau in the far distance sits a huge player built castle of stone brick and glass, torches glowing along its walls. Square white clouds sit high overhead in a flat blue sky with a square sun. Distant terrain fades into Minecraft's characteristic render distance fog.
As the door opens, flat daylight spills through the doorway. The person steps forward through the doorway onto the grass blocks and begins to run freely and naturally across the world toward the ravine, relaxed and joyful, natural human running motion, consistent body proportions and identity held from the photo.
The person stays completely photoreal. Real skin, real hair, real fabric, real human anatomy, exactly as in the attached photo. They are never stylized, never blocky, never pixelated. The contrast between the photoreal person and the fully blocky world is the point of the shot, and the boundary is a clean hard cut at the door frame with real interior on one side and Minecraft on the other.
The world is alive with motion authentic to the game: the square clouds drift slowly in one straight line at a constant speed, water and lava animate downward in looping texture frames, torch flames flicker with tiny floating embers, and sheep, cows and chickens wander through the grass with stiff blocky walk cycles. All motion is slow, steady and mechanical, exactly like the real game, nothing organic or wind driven.
Lighting is Minecraft's flat block lighting, each cube face evenly lit as a single tone, hard edged shadows aligned to the block grid, no soft falloff, no bloom, no ambient occlusion gradients, no lens flare. The person is lit to match that same flat daylight once through the doorway, with a hard edged shadow on the grass blocks beneath them, shadow direction consistent with the square sun.
No transition effects of any kind, no glowing edges, no flashes, no sparkles, no portal particles, no warping or morphing at the door frame. The world is rendered exactly like vanilla Minecraft gameplay footage, not a realistic or ray traced or cinematic reinterpretation of it. Static locked camera throughout.Step 5: Stitch the halves together
You now have two clips: your real footage, and the generated continuation. Upload both to the same chat and paste this.
Here are two clips. The first is my original footage, the second is the generated continuation.
Stitch them into one seamless video. Cut the first clip at the exact frame the door starts to open and join the second clip there. Match the color and exposure across the join so the cut is invisible, keep the original audio from the first clip running underneath, and export it vertical 9:16 at the highest quality available.Make it any world you want
The prompt is built so the world is swappable. Everything from the third paragraph onward describes Minecraft, so replace that block and keep the rest.
Some that work well: an underwater trench with whales drifting past, a neon Tokyo alley in the rain, the surface of Mars with the Earth low on the horizon, a medieval castle courtyard mid tournament, or a cloud city above a storm.
Two things to keep whatever world you pick. Keep the locked camera, because it's the anchor that sells the join. And keep the instruction that you stay photoreal, because the model will otherwise start restyling you to match the world as you cross the threshold, and the whole effect collapses the moment your face changes.
The things that break it
A moving camera in the original. Even slight handheld drift means the generated half cannot match, and the join reads as a cut instead of one continuous shot.
A motion blurred start frame. The model reads your face from that single still. Blur it and you get a stranger.
Asking for effects at the doorway. Glows, sparkles and portal shimmer sound good and look like AI. The prompt explicitly bans them so the boundary stays a clean hard edge, which is far more convincing.
Cutting at the wrong frame when stitching. Join at the exact frame the door starts moving. A few frames early or late and you get a visible jump in the door's position.
Get the next one first
New prompts every week.
Free. The new drops and the tools behind them, before they hit the feed.
No spam · New issues Sunday · Unsubscribe anytime
Need it custom?
Want this built for you?
Tell me the idea and I’ll build it. An app, a tool, an automation. You don’t need to be technical.
Related guides
Guide
The ChatGPT Photoshoot: Studio Headshots of Your Own Face (Higgsfield Plugin)
Lock your face as a character inside ChatGPT, then recreate any photo you find on Pinterest with it.
Prompt
Turn One Photo of Yourself Into a Cinema-Grade Anime (Claude + Higgsfield MCP)
The exact 3-step pipeline prompt that turns a single selfie into a consistent anime protagonist, a sidekick, and a 12-second cup-final scene, no drawing, no timeline.
Guide
The Higgsfield Prompt Vault: Every Workflow I Use, One Page
Every Higgsfield workflow from my videos in one place, the living website, the anime film, street interviews, drone flythroughs, plus every video prompt, copy-paste ready.
Prompt
The Cinematic 'Meet Your Future Self' Video (My Exact Prompt)
The exact prompt behind the cinematic two-selves edit: your younger self and your stronger future self, lit like a boxing ring at midnight. Drop in two photos, paste this, done.
Frequently asked questions
No. A phone on a tripod or propped against something solid is fine. The only real requirement is that it does not move at all while you film.
Seedance 2.5, through the Higgsfield plugin in ChatGPT. Get the plugin here, then select Seedance 2.5 when you send the start frame and prompt.
Almost always a soft or motion blurred start frame. Pick a sharper frame where your face is clearly readable. Keeping the line about staying photoreal in the prompt also stops the model restyling you to match the world.
Yes, that is the point of the structure. Swap the paragraphs describing the world and leave the camera, identity and no-effects instructions exactly as they are.
Three or four seconds. You only need the approach and the moment the door starts to open, since everything after that is generated.
