When one of these pictures looks fake, the giveaway is rarely the faces. It is the floor. Two copies of a person can be rendered well and still read as a paste-up because one of them has a long shadow and the other has none, or because their shadows point in different directions, which is a physical claim that two lights exist in a room that has one.
So the three ways of doing this are really three levels of difficulty for one problem, and they are worth doing in that order. Motion trails put copies at the same size and the same distance, so one light covers them all. Standing clones spread copies through the room at different depths, so each shadow has to be a different length. Scale contrast makes one copy six meters tall, which is the hardest shadow in the set. All three were run at 16:9 in CapCut from one photograph.
The source: one man, one hard light at the upper left, one unambiguous shadow. Everything below is this photograph with copies added.
Trails: the copies that share one light
The easiest version is the one where the copies are the same size and stand where the original has just been. The photograph was animated through the AI video generator flow at Seedance 2.0 Mini, 16:9, five seconds and 720p, with the trail written as a count and a fade.
The camera does not move. The man walks steadily from the left of the hall toward the right, and behind him he leaves a train of five fading afterimages of himself in the positions he has just passed through, each one more transparent than the one in front of it, all of them lit from the same hard light high on the left and each casting its own faint shadow to the right on the floor.
The finished clip, 1280 by 720, 24 frames per second, 5.04 seconds, as generated. Shown here as frames in sequence; the full video and a GIF version are supplied alongside this document.
The trail builds one copy at a time. By the last second there are five afterimages, and each has kept a shadow.
This works because the request is easy to satisfy: the copies are the same height, they stand on the same line, and their shadows are therefore near-identical and can be dimmed together as the copies fade. Nothing about the geometry has to be recalculated per copy, which is exactly what the harder versions ask for.
What "show him three times" does instead
The still version starts with the request most people would write, which is a count and nothing else.
Show the same man three times in this room.
The same photograph and the same count, requested two ways. The lower row is the floor of each at full size.
What came back is not three copies of the photograph, it is three different scenes: one copy sits on a chair that was not there, one stands at a window that was not there either, and one reads a book. The room was re-staged rather than populated, and the shadows follow the staging rather than the lamp.
The second version keeps the count and adds the light, the shadow direction, the depth of each copy and one overlap.
Keep the single hard light high on the left: every copy is lit on its left side and in shadow on its right, and every copy throws its own sharp-edged shadow across the floor to its right, at the same angle as the original shadow and getting shorter for the copies that stand further from the camera.
That version returned three standing copies with their shadows running into the same quadrant of the floor. Measuring the direction of the darkest ray leaving each figure's feet, the source photograph reads 24 degrees below horizontal to the right, and the three copies read 24, 48 and 3 degrees, so all of them fall to the lower right and none of them contradicts the lamp. The figures were located with a standard person detector and the floor in these generations is heavily stained, so treat those numbers as a direction check rather than as survey data.
Six copies, and what gives way
The same instruction was then asked for six copies instead of two, with everything else identical.
Six copies and the scale version. The enlarged panels show what happened at the front of one and at the feet of the other.
The direction survived and the length did not. The two copies nearest the camera share one long shadow reaching most of the way across the floor, while the four standing further back have short dark pools directly under them, as though the light had moved up to the ceiling for them alone. There is also a pair at the left that overlaps into a single silhouette with three legs, which is the kind of thing that is invisible at thumbnail size and obvious at full size.
Scale contrast is the hardest shadow in the set
The third route asks for a six-meter copy and a twenty-centimeter copy in the same hall as the original. The two small figures came back well: the miniature has a small sharp shadow of its own and the original keeps the long one it started with.
The giant has no shadow at all. Its feet meet a patch of clean lit floor, and the space where a six-meter shadow would fall, which would be the largest dark shape in the picture, is empty. That is the single most expensive omission available in this effect, because scale contrast works by making the viewer compare sizes, and a shadow is the main thing that tells them a large object is standing on the same floor rather than floating in front of it.
What the six requests cost
The two cost lines with the settings rows they belong to, read on 10 September 2026.
Thirty credits, twenty-five of them for the one moving version. Worth noting for anyone budgeting a series of these: the trail is the expensive request and the one that needed the least correcting, while the still versions are cheap enough to run twice, which is what the comparison above cost. These figures are for these configurations on this date and come from the composer's own cost indicator; other durations, models, resolutions and modes are charged differently.
Where to look when you check it
Check the floor first and the faces last. Go to each copy's feet and ask three questions in order: is there a shadow at all, does it leave the feet on the same side as every other shadow in the frame, and is its length reasonable for how far away that copy is standing. In this run the first question caught the giant, the second was fine everywhere once the light was written into the request, and the third is what broke at six copies.
When one copy fails, re-send the whole image with that copy described by position rather than editing the picture, since the copies are generated together and a request that names one of them by where it stands is the cheapest way to move it. And expect to lose some of the room: every version here re-rendered the concrete, so the clean walls of the source came back stained in all four, even though the request said the hall stays exactly as it is. If the room matters as much as the copies, that is a reason to keep the background plain enough that a re-render cannot spoil it.
The man in these pictures is generated. With a real photograph, this effect makes several copies of a specific person doing things they did not do, which is worth keeping to your own face or to someone who has agreed. The setting here is a bare invented hall on purpose, and locations for this kind of shot are better described generically than named, so that a copy of somebody does not end up standing somewhere they have not been. What this article does not cover: cutting the trail to a beat, background removal, repainting a single copy by hand, upscaling and 4K export.
Written 10 September 2026. The source photograph, all four clone versions and the clip were generated in CapCut for this article; the man is invented and no real person, place, product or brand appears. Interface labels and the cost lines reflect a single session on that date and the configurations named above, and may change. Shadow directions were measured by taking the darkest ray leaving each figure's feet, with the figures located by a standard person detector.
内部审校参考（发布前删除此节）
锚定玩法：男子卡点残影特效（ID 7，M1 单图人像特效，P0，互动热度 154）；其他锚定玩法 潮男超现实空间分身(h3)、微缩迷你人分身(h3)、女孩巨大化(h105) 三条并入"分身"与"尺度反差"两节，均已做出。标题与 slug 与规划行一致。
规划行"内容结构建议"中的卡点剪辑、局部重绘、背景移除、画质增强、4K 导出本次未做，已在文末范围说明列出。
风险与备注"不要指向具体世界遗产地或真实景区的可识别标志物"：本篇场景为虚构空混凝土厅，末节另写明地点应作通用描述。
核对口径（2026-09-10 生产前核验 TikTok）：
"clone yourself same photo ai effect" → 该词最高赞 45.2K（How to clone yourself）、14.7K（#clone 摆拍）、4146（I cloned myself with AI in 20 minutes）、787、147；
"motion trail echo effect edit" → 最高赞 33.2K（echo motion blur edit）、6576、5107，另有一条 CapCut 幽灵残影教程；
"miniature and giant person ai photo" → 最高赞 14.9K（AI Giant Statue Filter，带 CapCut 入口）、332（#capcutpioneer #capcutnow，3 天前）。三条路线均在更新。
clone yourself same photo ai effect；
motion trail echo effect edit