Keep Feed Icons Still While a Character Moves Between Posts

Separate a cross-post image into matched background and transparent character exports, then animate the foreground in CapCut without moving the feed interface.

A sheet ghost and an adult in an orange jacket move right and return while the heart, comment and bookmark icons remain fixed between two photographs.
CapCut
CapCut
Oct 28, 2026

You have a Halloween image in which a ghost reaches from one post into another, and you want a short movement without dragging the feed icons along. This CapCut workflow turns that still into a five-second foreground sway while the account text and two photo backgrounds stay in place.

The example uses an original AI-generated adult and a generic sheet ghost. We tested the web workflow on October 21, 2026. Both characters move together as one cutout; the hand keeps its grip on the jacket. This is a flat character layer moving sideways, rather than a newly generated lifting action or articulated body animation. The posts are part of a graphic, not a working social interface.

A sheet ghost and an adult in an orange jacket move right and return while the heart, comment and bookmark icons remain fixed between two photographs.

Four frames from the same accepted native motion export, read across each row at 0.000, 1.233, 2.500 and 4.967 seconds. The HTML edition shows the movement as a GIF. Native AI indicators are retained.

Keep feed icons on the layer that stays put

The important split is between the characters and everything behind them. In this image, the ghost's arm crosses a charcoal interface band to hold the adult's jacket. Separating those two characters into different moving pieces would add a second alignment problem at the hand. We kept their connection inside one foreground image.

1. Open CapCut AI Design and start a new design. Upload your accepted cross-post still. Our source was a 2048 × 1152 landscape PNG from the preceding static-image example, with fictional account names and no real person's photo.

2. Select the uploaded image and choose Separate layers. After the operation completes, open Layers and expand Frame 1. Our run produced four editable text layers and two IMAGE layers: a connected ghost-and-adult cutout, and a background containing the photo settings, icons and avatars.

Actual CapCut Layers panel showing four text entries and two IMAGE entries inside Frame 1, with the two image rows outlined in amber.

The separated frame in this run. The upper IMAGE row contains both characters; the lower IMAGE row contains the background and interface graphics. The uploaded original is listed outside the frame.

Hide the character layer to examine the background before animating. In our exported background, the old character areas had been filled in. That matters when the foreground moves away from its starting position: a copy of the character left on the background would appear behind the moving cutout. Check the vacated area on your own result before continuing.

This separation also changed the typography into editable text. Treat the separated frame as a new version to inspect; it is not a pixel-preserved copy of the uploaded still.

Export two aligned images from the separated design

3. Make the background IMAGE and the four text layers visible, with the character IMAGE hidden. Select Frame 1, then use Download to export that frame as PNG at 1× (2048 × 1152px). Leave the uploaded original unchecked in the export list.

Read the downloaded names and caption at full size. Our first background export clipped some of the separated text. We reduced the text layers' Font size in the design until the full labels fitted, then exported the background again. In the accepted version, both account names and the complete caption fit inside their text areas. Check the saved PNG, since the small canvas preview did not reveal the clipping clearly.

4. Hide the background IMAGE and the four text layers, then show the connected character IMAGE. Select Frame 1 again. Download a second PNG with Transparent background enabled, using the same 1× size.

Native Download panel with Frame 1 checked, the uploaded original unchecked, PNG at 2048 by 1152 and Transparent background enabled.

Exporting the whole frame preserves the cutout's position within a 2048 × 1152 transparent canvas. The amber outlines mark the frame selection and size/transparency controls.

Our individual character-layer download measured 1066 × 1103, while the full-frame transparent export measured 2048 × 1152. We used the full-frame version in the video editor so that both inputs shared the same canvas coordinates. The empty transparent space is intentional; do not trim it from this matched pair.

Move the connected characters across one fixed band

5. Create a project in the CapCut online video editor and upload the two accepted PNGs. Add the background to the main track, then drag the transparent character image onto a track above it. In our project, both image clips lasted five seconds and started at the same timeline position.

Select the upper character clip and open Basic. Under Transform, our starting values were Scale 100%, Position X 0, Y 0 and Rotate 0°. Keep the background clip's transform unchanged while you set the foreground path.

6. Move the playhead to the start and add a Position keyframe with the diamond beside the controls. Move to the midpoint and enter X 60; move to the last visible frame and enter X 0. Changing X at the later positions added the next keyframes in this run.

Playhead at 30 fps
Foreground X / Y
Purpose in this example
00:00:00
0 / 0
Begin at the aligned source position.
00:02:15
60 / 0
Shift the connected characters to the right.
00:04:29
0 / 0
Return on the last picture frame.
Actual CapCut Transform controls at the midpoint showing Scale 100 percent, Position X 60 and Y 0, Rotate 0 degrees and an active Position keyframe diamond.

The midpoint values on the selected foreground clip. The amber outline marks Position and its keyframe control; the background track has no position movement.

The ghost's torso ends at the upper photo edge in this cutout. We kept Y at 0 and used a small horizontal movement so that this flat edge stayed on the same horizontal boundary. The ghost's arm and the jacket grip traveled together. For this composition, that was a useful direction to test before attempting a larger movement that could expose more of the cut edge.

Compare the icons at the start, midpoint and return

7. Choose Export, then Download. Our accepted export used 1080p, Recommended quality, 30fps and MP4. The saved picture stream measured 1920 × 1080 with 150 frames over five seconds.

Native video export settings showing 1080p, Recommended quality, 30fps and MP4, with the settings outlined in amber.

The export settings used for the tested motion. These are the selections observed in this project, rather than a list of available product specifications.

Compare the outline icons against the frame edges as the character moves. We checked the heart, comment and bookmark regions plus the account text across all 150 native frames. Position matching found no displacement in those measured regions. There were small compression differences between frames, so this does not mean the exported pixels were identical.

At 2.5 seconds, the ghost's face was about 60 pixels to the right of its first-frame position; at the last frame it was back at the start. In the inspected start, intermediate, midpoint and return frames, the hand remained connected to the jacket and no duplicate character was visible behind the moving layer. Both photo boundaries stayed horizontal.

Before using your export, check three things on the saved file: the full account labels, the character's vacated starting area, and the hand-to-clothing contact at the midpoint. If those pass, keep the matched PNG pair with that accepted export so you can adjust the motion without rebuilding the still.

Hot and trending