What is an AI real estate video?
An AI real estate video is a listing clip built from photographs instead of footage. You take a still you already have, a wide living room shot, a front elevation at dusk, an aerial of the lot, and an image-to-video model animates it into a short clip by pushing the camera slowly forward into the scene. Nothing in the property changes. The house stays the house; the camera is what moves, and it only ever moves deeper into what the photograph already showed. Done well, the result is indistinguishable from a motion-controlled slider shot, which is exactly what a listing video is mostly made of.
Two different things are called "AI real estate video"
Almost every piece of confusion in this space comes from conflating them, and no article we found separates them clearly. The first is motion from stills: your real photo, with depth-based camera movement added. The pixels stay real, nothing is invented, and the only thing that changes is where the camera is. Because those pixels came out of edited HDR stills, you often inherit better colour, better lighting and straighter horizons than you would get from actually filming the house. The second is generative transformation: taking a real photo and generating a new image of what the space could be, then animating the change. Virtual staging, an unfinished basement becoming a living room, a bare lot becoming a build.
They are different products with different buyers and different ethics. The first is a production shortcut; the second is a service line you sell as "here is what this could become." Everything on this page is the first kind, deliberately, and taken to its strictest form: the source photo is treated as the hard boundary of the scene, so the clip stays a record of the property rather than an interpretation of it.
This page gives you the prompt, not the render
Every model that does this well is prompt-hungry, and that is where most agents lose. Type "make this photo move" and you get a wobbling room with a bowed wall. The prompt has to specify the exact camera move, the speed, the axis, what may move (curtains, water, cloud, light) and, most importantly, a long list of what may not, including an explicit ban on outpainting. Each entry on this page is that finished prompt, already written and tested against real listing photos. Copy it into whichever AI video model you already pay for, or hit Create video and it opens pre-filled in the AI video studio we link to. You supply the photo of the actual property; we supply the language.
Which camera move belongs in which room
Almost none of them, and that is the most useful thing on this page. The instinct is to reach for a lateral track down the kitchen island, an orbit around the pool, a crane rise in the bedroom, an aerial that pulls back over the roof. Every one of those breaks, and for the same structural reason: they all force the camera to reveal area the photograph never contained. Pan right and pixels have to appear on the left out of nowhere. Pull back and the frame grows at all four edges at once. The model cannot refuse, so it guesses, and the guesses are the invented walls, phantom doorways and extra rooms you have seen in bad AI listing clips.
Forward is the one direction that is safe, because moving deeper into a scene crops the frame inward. Every frame a push-in produces is built from pixels the photo already had. So the whole library is push-ins, and the variety comes from what you aim them at: the centre of a living room, a point down the kitchen island so the shot reads with a drift, the window in a bedroom, the far end of a pool. The one shot that seems impossible under this rule, the aerial, is solved by reversing it: a descent toward the house crops inward exactly like a push, so you get the drone beat without the invention
Why AI interiors warp, and how these prompts stop it
Two things go wrong, and they get blamed on each other. The first is deformation: architecture is unforgiving, so where a person's arm can bend a few degrees wrong and nobody notices, a wall that bows by two degrees is instantly fake. The second, and by far the bigger one, is invention. Any sideways or backward camera move requires pixels the photograph never contained, so the model fills them in and you get doorways and furniture that do not exist. Every prompt here closes both: the camera is restricted to a forward push so the frame can only crop inward, outpainting is banned outright, and the architecture lock spells out that walls, window frames, cabinet lines, countertops and rooflines stay straight and dimensionally accurate with nothing added, removed or restyled. The instruction that does the most work is the last one: when in doubt, preserve the original pixels instead of generating new ones, because a nearly static shot always beats a hallucinated one
The counterintuitive part: stills can beat filming
The strongest argument for this whole approach is not that it is cheap. It is that the source material is better. The only way to film an interior with correct exposure both indoors and out through the window is to flood the house with light, which in practice means a truck of gear and a full crew. Effectively no listing videographer has that time or budget, so windows blow out and you lose the view. A photograph does not have that problem, because a photograph can be edited: you can recover a blown-out window so the view is visible, and you can clean up the wall. You cannot do either to footage. So you edit the stills, then generate the video from the edited stills, and the result can reach a quality the filmed version was never going to. That argument comes from a photographer who went into testing this as a skeptic and came out calling it the future of the category.
What this replaces, and what it doesn't
A full listing video package with drone coverage typically runs from several hundred dollars to well past a thousand per property, and the twilight exterior alone usually needs a second site visit inside a fifteen-minute window with every light in the house switched on. Animating photos you already own covers most of that shot list at no marginal cost per listing, which changes the maths for the mid-market properties that never justified a video crew. What it does not replace is a real walkthrough: an image-to-video model cannot invent the hallway behind a doorway it never saw, so it moves the camera within a scene rather than through the house. One important line to hold: keep the prompts to camera movement and ambient atmospherics. Using AI to add furniture, change finishes or alter a view crosses from cinematography into misrepresentation, and your MLS rules on enhanced media apply exactly as they do to photos.
The shot library
Six finished property shots, each with the complete prompt that produced it. Copy it, or open it pre-filled and upload your own listing photo.
Interior
Living Room Push-In
The slow push-in is the most-used shot in luxury real estate video and the only camera move that is genuinely safe to generate from a still. The reason is geometric, not stylistic: pushing forward crops the frame inward, so every frame it produces is made of pixels that were already in your photo. Pan sideways instead and something has to appear at the edge that the photo never contained, which is where invented walls, doors and furniture come from. This template takes a standard wide living room photo, the kind every listing already has, and creeps the camera forward at a motion-controlled crawl while sheer curtains breathe and dust drifts through the window light. Use it as the opening beat of a listing reel.
Input photo: A wide, straight-on living room photo
Kitchen Push-In
Kitchens sell houses, and a still flattens them: buyers cannot tell whether an island is six feet or twelve. The instinct is to track the camera sideways down the counter, and that is exactly the move that breaks. Sliding sideways forces the model to invent whatever sits past the edge of your photo, which is how you end up with a cabinet run that never existed. The fix is an off-center push: aim the forward dolly at a point down the island rather than at the center of the frame. The frame still only ever crops inward, so nothing is invented, but because the target is off to one side the shot reads with a gentle drift and the near and far ends of the island separate enough to give you real depth. It is the closest thing to a lateral move that does not ask the model to make anything up.
Input photo: A kitchen photo along the island
Primary Suite Push-In
When the view out of the window is worth more than the room, a photo makes you choose which one to feature. The usual answer is a crane rise with a downward tilt, and it is the wrong answer here: a tilt swings the camera's viewing direction, which drags unseen ceiling or floor into frame and hands the model a blank area to invent. A forward push aimed at the window gets you the same story with none of the risk. The shot begins wide enough to read the whole room, then crops inward until the window and the view carry the frame, and every frame along the way is built from pixels the photo already had. Linen and curtains get the only real motion, which is enough to keep it alive. Use it as the last interior beat before you cut outside.
Input photo: A bedroom photo with a window in frame
Exterior
Twilight Facade
Twilight exteriors are the most requested and most expensive shot in real estate photography, because a photographer has to be back on site inside a fifteen-minute window with every interior light switched on. Animating a front elevation you already have gets you the moving version without the second visit. The camera pushes slowly toward the facade, cropping inward the whole time, while the sky deepens toward blue hour and the windows warm up. That lighting shift is the one sanctioned exception to the image lock on this page, and it is safe for a specific reason: changing ambient light does not change geometry, so it never asks the model to invent anything. The roofline, window frames, siding courses and driveway all stay exactly as photographed, and no lamp or fixture that is not already in the picture is allowed to appear.
Input photo: A front elevation, daylight or dusk
Pool & Patio Push-In
Outdoor space is the amenity buyers scroll for, and the obvious move is an orbit around the pool. It is also the single most destructive move you can ask for from a still, because swinging the camera around an object demands the far side of everything, and the photo only ever contained one side. What you get back is a reshaped deck and a house that quietly rebuilds itself. A forward push across the water gives you the space without the invention: the frame crops inward toward the far end of the pool, the deck and the back elevation grow in scale, and the water gets the one thing a photo can never show, a continuous gentle ripple with light moving on the surface. Everything solid stays locked, so the only things moving are water, foliage and the camera.
Input photo: A pool photo with the house behind
Garden Approach
Curb appeal is not only the front door. On a property where the grounds are part of what is being sold, the shot that does the work is the approach: the camera moving through the garden toward the house, so the planting, the lawn and the terrace pass through frame on the way in and the buyer gets a sense of walking the plot rather than looking at a picture of it. It is also one of the most forgiving shots to generate, because a garden photo usually has real depth in front of the house, which is exactly the material a forward push needs. Foliage supplies the ambient motion, the house holds still, and because the camera only ever moves deeper into the frame, nothing beyond the edges of your photograph is ever required. Use it to open an exterior sequence or to close the tour.
Input photo: A garden or grounds photo with the house in view
The motion templates
Six moves written against your own photo. Every one starts from a single uploaded listing image and treats it as the hard boundary of the scene.
Hero Push-In
If you only ever use one real estate video prompt, use this one. A slow forward push works on almost any interior photo and it is the only move that is structurally safe from a still: travelling forward crops the frame inward, so every frame is assembled from pixels the photograph already contained. Nothing has to be guessed, so nothing gets invented. Pair it with a wide, straight-on room photo for the strongest result.
Off-Center Push
This is the template to reach for when you want the feel of a sideways tracking move, which is how you show a buyer that a space has real length. A true lateral track is the move that breaks worst from a still, because sliding sideways forces the model to invent whatever sits past the edge of the photo. The off-center push gets most of the effect for none of the risk: you keep the dolly purely forward, but aim it at a point off to one side of frame. The shot drifts in that direction, near and far objects separate enough to read as depth, and the frame still only ever crops inward.
Day to Twilight
Twilight exteriors are the single most expensive shot in a listing photography package, because they need a second site visit inside a fifteen-minute window with every light in the house on. This template gets the same effect out of an ordinary daylight exterior: the sky deepens toward blue hour and the windows warm up over a slow forward push. It is the one place on this page where the source image is allowed to change, and it is safe for a precise reason: shifting ambient light never changes geometry, so it never asks the model to invent anything. No lamp or fixture that is not already in your photo is permitted to appear.
Window Push
When the view is the selling point, you want a shot that starts on the room and ends on the glass. The conventional way is a crane rise with a downward tilt, and a tilt is a trap from a still image: swinging the viewing direction pulls unseen ceiling or floor into frame and hands the model a blank region to fill in. Aiming a straight forward push at the window does the same storytelling job with the camera direction locked, so the frame only crops inward and the view stays the view that was in your photograph rather than a re-rendered guess at it.
Aerial Descent
Every listing video wants the aerial that rises away from the roof to reveal the neighbourhood, and it is the one shot you cannot generate from a still. Retreating makes the frame grow, so the model has to invent streets, roofs and trees at every edge for the entire clip. Reversing the move solves it completely: a descent toward the property crops inward instead, the house grows in frame, and nothing outside the photograph is ever required. It reads as a drone settling onto a property rather than lifting off one, which works just as well as an opener.
Vertical Listing Reel
Most listing video advice still assumes a 16:9 YouTube edit, but the clips that actually generate inquiries now live in Reels, TikTok and Shorts, where a horizontal crop wastes most of the screen. This template runs the same reliable forward push, reframed vertically. One important detail is built into the prompt: the 9:16 frame must be produced by cropping into the source photograph, never by generating new image area above or below it, which is the shortcut that fills a vertical frame with invented ceiling and floor. Stack three or four of these from different rooms and you have a listing reel.