AI camera control: a new camera angle for your photo

Rooms, houses, people, products, food, pets, landscapes, and streets; ordinary phone photos are the normal input and every example on this page started as one. Photos with one clear scene and some depth (a room with a far wall, a street, a person in a place) give the most convincing moves. A flat wall, a screenshot, or a tightly cropped object gives the tool little geometry to work with.

The same scene at the same instant photographed from somewhere else. The people, their outfits, the furniture, the light, and the label text stay; the camera's position, height, and distance change, so things at different depths shift against each other and new sides come into view. It is not a crop, a zoom, or a mirror.

The tool picks one bold move that a photographer would take for that kind of photo: a 45-degree orbit for a person, the opposite end of a room looking back, a step to the side of a house, a walk up the path in a landscape, an orbit around a pet at its eye level. It never changes the season, the light, or anything in the scene.

Yes. Source-visible words, numbers, and logos are copied character for character and shown from the new angle, and no lettering is added to anything that had none. Ask for a move that keeps the label facing the camera, such as “orbit right and come closer, keeping the label readable”. Small print that is already unreadable in the upload stays soft rather than being invented.

Yes: “from the doorway”, “from the window side”, “from behind the counter”, “from the far corner by the bookshelf”. The camera stands at that spot at eye height and looks back at the subject. The living-room example at the top of this page was made that way.

Yes: “from the floor, looking up”, “low angle”, “from above, high angle”, “overhead”, “bird's-eye view”. The horizon and the vertical lines answer the new height. For a room, an overhead view needs a ceiling the photo does not show, so a high corner is more honest than a straight top-down view.

Rooms and houses for listings, when you have one photo of a room and want it from the other end; products, to show the other side or a low three-quarter view without a reshoot; portraits, for a low or high angle of the same moment; storyboards and previsualisation, to try a shot before you move a real camera.

An orbit turns the camera around the subject by 35 to 90 degrees; closer halves the distance or more; further doubles it or more; higher and lower are a clear change of height, from the floor to overhead. A named spot means the camera stands there. The contract treats a result that is hard to tell apart from the upload as a failure, so it never returns a slightly shifted crop.

Every wall, window, door, floor, and object keeps its size and place, and every person keeps their face, outfit, pose, and gaze, as if a second camera had fired at the same instant. What the move reveals is continued plainly from what is visible: the same brick around the corner, the same floor under the camera, a bare counter. Nothing is added, removed, moved, or restyled.

Yes. Every example on this page is the tool's own output, made from the upload beside it with the direction printed under it, and nothing was retouched afterwards. The board pairs use plain phone photos; the use cases add 100% crops of a face, a label, and a dog's eyes so you can check that the detail survived the move.

4:5 portrait by default, or any supported ratio you choose under Settings before you create. When the ratio differs from the upload, the tool extends the scene from the evidence rather than cropping or stretching it.

A move that has to reveal a lot the photo never showed: the back of a house, the room behind the camera, a face turned fully away. The tool continues those surfaces plainly, which is honest but can look bare, and it sometimes lands a shorter orbit than asked. Ask for the move in one sentence, name a spot or a height, and run it again if the first result is too timid.