A football edit that flips into anime style is two clips and one join. The join is where these edits come apart: the drawing changes and the action jumps at the same moment, so the switch reads as a mistake instead of a deliberate change. This article builds one at 16:9 in CapCut, puts the cut on the frame where the boot meets the ball, and then breaks the same edit on purpose by moving that cut fifteen frames earlier.
The finished edit, 4.9 seconds at 16:9. Live action for 2.7 seconds, then anime from the frame of contact onward. Shown here as frames in sequence; the full video and a GIF version are supplied alongside this document.
A football switch is a cut between two clips that share one frame
The rule behind a clean switch is short: the anime half has to begin on a redrawn copy of the live half's last frame. Not a fresh drawing of the same player in roughly the same position. That frame, redrawn. When both halves hold the same body in the same place at the same instant, the viewer reads the join as a change of drawing and the action carries straight across it. Everything below serves that single frame: making a live clip that contains a usable one, finding it, redrawing it, and animating forward from the drawing.
The football here was generated rather than filmed. Broadcast and stadium footage belongs to whoever shot it, so this session generated a stand-in instead: one player in a plain red kit with no crest, no sponsor text and no number, on a floodlit pitch. If you already have a clip that is yours to use, the generation steps are optional and the method starts at finding the switch frame. Sound design and music are not covered here, and the anime half is a generated clip rather than hand-drawn animation.
Build the live half first, and give the strike room to happen
The first attempt at the live clip started from a still of the player already loaded for the shot, standing leg planted, ball at his foot. The clip that came back put contact at 0.71 seconds, seventeen frames in. Since the anime half takes over from the moment of contact, that left almost no live footage to establish anything. Where the strike lands inside the generated clip sets the length of the live half, and the source still is what moves it.
The second attempt started three strides earlier. The still puts the player mid-stride on the left third and the ball on the right third, with open grass between them, so the clip has somewhere to run before it arrives.
Photoreal wide action photograph of a football striker running up to shoot. Low camera angle close to the grass, side-on view. He is in his mid twenties with short dark hair, wearing a plain deep red jersey with a white collar, plain white shorts and red socks, no crest, no sponsor text and no numbers on the kit. He is three strides away from a plain white ball that sits on the grass ahead of him, both feet off the ground mid-stride, arms out for balance, eyes on the ball. Night match under floodlights, blurred crowd far behind, cool rim light on his shoulders. Sharp focus on the player, shallow depth of field. 16:9 landscape, the player on the left third and the ball on the right third with open grass between them.
The starting still for the live half. The gap between player and ball is the run-up the clip needs.
That still became the first frame of the video request. The prompt describes the approach, the strike and the follow-through in order, and pins the camera, because a camera that moves during the live half makes the geometry harder to match later.
He takes two more strides, plants his left foot beside the ball and strikes it with his right boot. The ball leaves the grass and travels away to the right of frame while he follows through. The camera stays fixed at the same low angle near the grass, no zoom, no pan, no cut. Photoreal live action, floodlit night stadium, blurred crowd behind him.
Both clips were generated with the CapCut AI video generator in Video Studio. In the web interface used for this build, attaching the still to the home composer and switching the mode menu to Video opened a settings chip. The panel showed Video clip and Full video, Model and quality, Aspect ratio, and Duration controls. This run used Video clip, Seedance 2.0 Mini, 720p, 16:9 and 5s. Check these settings before each request because available controls and defaults can vary by version, region and account.
The panel behind the composer chip. Aspect ratio and duration are set here, before the send.
The interface showed an estimated credit cost before each request ran. Credit use can vary by model, duration, resolution, mode, account and region, so check the current estimate shown in your version before generating rather than carrying a previous session's numbers into a new project.
The settings line and the price line, both shown in the chat before the clip is generated.
The clip came back at 1280 by 720, 24 frames per second, a little over five seconds, in about four minutes. It also drifted: the prompt asked for a locked camera and the framing tightened on the player as he ran. That drift does not matter for the join, because the anime half is built from a frame of this clip and inherits whatever framing that frame has.
Two source stills, two very different edits. Contact at 0.71 seconds leaves seventeen frames of live footage; contact at 2.71 seconds leaves sixty-five.
Find the frame where the ball leaves the boot
A strike offers three moments that could carry a switch: the plant of the standing foot, the contact, and the landing. Contact is the one to use, because it is the only one where something in the picture changes state. The ball is on the grass, then it is in the air. A viewer who is watching the ball is busy at that instant, and a change of drawing arrives while their attention is on the thing that just moved.
Finding it means stepping through the clip rather than guessing. This one runs at 24 frames per second: the boot reaches the ball at frame 65, at 2.71 seconds, and by frame 66 the ball has left the boot with motion blur trailing it. Frame 66 is the switch frame. In the CapCut web editor used for this build, the position readout displayed minutes, seconds and frames, so the same moment could be found by scrubbing and reading the numbers, and the clip can be downloaded from the canvas card for a closer look.
Redraw that frame instead of drawing a new one
The switch frame went back into Video Studio as a reference attachment in Image mode, using the same AI image generator flow that made the source still, with a prompt that asks for a redraw rather than a new picture. The instruction to hold the camera angle, the body position, the ball position and the framing is the part doing the work; the style words follow it.
Redraw this photo as a single frame of a Japanese sports animation. Keep the camera angle, the player's body position, the position of the ball and the framing unchanged. Bold black ink outlines, flat cel shading with hard-edged shadows, a simplified crowd painted as blocks of color, the same deep red jersey with a white collar, white shorts and red socks. Draw speed lines behind the ball and a warm floodlight glow into the sky. Hand-drawn look with no photographic texture. 16:9 landscape.
The live frame and its redraw, with the boot and ball enlarged. The pose and the ball hold; the stands are rebuilt.
The redraw held the geometry. In this pass the ball came back within about forty pixels of its position on a frame 1280 pixels wide, the standing leg and the swinging boot line up, and the floodlights stay where they were. The background is where the two frames disagree: the crowd line sits higher and the colors are far stronger than in the photograph. That difference is visible at the join, and it is the price of getting the player right.
Asking for the speed lines and the floodlight glow inside this prompt matters for a reason worth naming. They are drawn as part of the picture, in the same pass as the player, so they sit behind and around him at the depth the drawing gives them. Effects added on top of a finished frame afterwards tend to look like a layer sitting on the picture, because that is what they are. Before generating, check the current credit estimate shown for the selected model, aspect ratio and resolution.
Start the second clip from the drawing
The drawing then goes back to the home composer as the attachment for a second video request, at the same Video clip, Seedance 2.0 Mini, 16:9, 5s and 720p. The attached image provides the visual starting point for the generated clip. Because generated motion can still alter framing or geometry, check the opening frames and use the closest match to carry the action forward.
The ball rockets away toward the right of frame trailing drawn speed lines while the player finishes his follow-through, his standing foot pivoting on the grass and his shirt moving. The drawn crowd stays still behind him. The camera stays fixed at the same low angle, no zoom, no pan, no cut. Hand-drawn Japanese sports animation, bold ink outlines, flat cel shading.
What came back uses its first second well and then runs out of action: the ball travels right and clears the frame by about one second in, after which the follow-through pose holds with small movements for the remaining four. Only the first 2.2 seconds went into the edit. That is worth planning for, since the useful part of an animated continuation may be much shorter than the clip you paid for. The interface displayed an estimated credit cost before generation; check the current estimate for your selected settings before running the request.
Join the halves on that frame
For this build, assembly happened in the web editor opened from Video Editor in the left rail. The live clip goes in first, dropped onto the preview area, which puts it on the timeline. Moving the playhead to 00:02:22, the timeline's reading of the contact frame, and pressing Split cuts the clip there; deleting the second piece leaves the live half ending on contact. Dropping the anime clip onto the timeline places it at the playhead, on a track above, and dragging it down onto the main track leaves the two clips end to end. A second split at 00:04:28 and one more delete trims the anime tail.
The two halves meeting at 00:02:22. The preview at that position is the redrawn frame.
One detail to expect: the editor timeline ran at 30 frames per second while both clips were 24, so the join lands on the timeline's frame grid rather than the clip's. In this build that put the cut at 2.73 seconds, between the contact frame and the frame after it, which is where it was wanted. Export settings offered 720p, Recommended quality, 30fps and MP4, and the finished file came out at 4.93 seconds. The export itself needed patience: three attempts returned Couldn't save, and closing the dialog and starting a fresh Export, Download, Export produced the file.
Move the cut fifteen frames and watch it break
The same two clips, cut at 00:02:07 instead of 00:02:22, show the difference directly. Fifteen frames is half a second at this timeline's rate, and the switch now happens while the leg is still travelling.
The same halves cut half a second early. The style change lands, and the strike happens twice. Shown here as frames in sequence; the full video and a GIF version are supplied alongside this document.
Both joins, frame by frame. On contact the pose carries across; fifteen frames early it does not.
Read the bottom row. In the last live frame the kicking leg is behind the player and the ball is still on the grass; in the first anime frame the leg has already swung through and the ball has moved. The style change is identical in both versions, so the half second of missing action is the whole difference. A join that reads as a glitch is a timing problem before it is a rendering one.
What a white flash does to the join
In the web editor version used for this build, a transition could be applied at the join between two adjacent clips and its duration adjusted. White Flash was available in that session and suited the impact, with an initial duration of 0.5 seconds. Transition names, placement and availability can vary by version or region.
The transition attaches to the join, and its duration field decides how much of the edit it covers.
Exported and measured frame by frame, the flash in this run spanned 2.57 to 2.93 seconds, so it started about a sixth of a second before the join and finished about two tenths after it. It also flickered rather than rising and falling once: bright frames and normal frames alternated across that stretch, with the two brightest at 2.90 and 2.93 seconds.
The White Flash at its default duration, sampled across the join. The frames the two halves share are underneath the brightest part of it.
That is the trade to weigh. The flash hides the change of drawing, which is useful when the two backgrounds disagree as much as these do, and it also hides the matched frame that took the most work to get right. It began before contact in this run, so the whiteout starts while the ball is still on the grass. The version at the top of this article uses no transition for that reason. Whichever way you go, judge it the same way: step through the join one frame at a time, and check that the pose and the ball are in the same place on both sides of it before deciding what to lay on top.
Written 4 September 2026. The player, the pitch and both clips in this article were generated in CapCut during a single session and depict no real person, club or competition; the kit carries no crest, sponsor text or number. Interface labels, timings and credit figures reflect that session and the configurations named above, and may change. If you rebuild this with match footage, use footage you have the right to use.