A reflected character can stop feeling like a reflection surprisingly quickly. One ear changes shape, the lower figure drifts sideways, or the two copies begin doing different things. The interesting design problem is keeping a relationship intact.
For a free image-to-video study, start by making that relationship visible in the still artwork. Here, one paper fox sits above a horizontal line. A vertically inverted copy sits below it. The proposed change is a small sideways slide of the whole pair.
This is an original illustrative exercise made with AI-generated artwork and AI-assisted text drafting. The body images use the same fox source with external compositing; the cover is a separate illustration. No video has been generated or evaluated. The construction is a flat graphic convention, not a demonstration of real optics or animal movement.
Declare the axis before adding movement
The horizontal line is the rule of the composition. In the 1600×900 planning canvas, it sits halfway down, at y = 450. The upright fox occupies the upper half; its inverted counterpart occupies the lower half.
Flip the fox vertically across that line. Do not turn it around horizontally: its nose still points right in both halves. A point at y = 173 above the line corresponds to approximately y = 727 below it. Both points keep the same horizontal position.
The narrow, equal gap on either side of the line is deliberate graphic spacing. It is not a claim about a fox standing on a real reflective surface. Keeping the model this simple makes its intended relationship easier to inspect.
An original AI illustration, composed externally into a flat reflection study. The lower fox is a vertical inversion of the same source; this is a still plan, not a video frame or a real reflection photograph.
Preserve identity through correspondence
Use one source illustration for both figures. Independently generating a second “matching fox” could introduce different folds, eye placement, or tail proportions before any motion is attempted.
Choose one recognisable detail to compare: a cream ear fold, the sharp nose, or the edge of the tail tip. Its counterpart should have the same shape, inverted vertically. A temporary guide connecting the two points should run straight up and down.
The guide image marks one such correspondence. It is a construction aid, not part of the final visual effect. The lower layer is dimmer to separate the halves, but its silhouette comes from the same source.
This gives you two different questions to ask. Is it recognisably the same fox? Is it in the correct reflected position? A pleasant colour match alone cannot answer the second question.
The marked points share one horizontal coordinate and sit equally far from the horizontal axis. The construction preserves the source silhouette rather than redrawing a second fox.
Move the pair, then check the margins
Prepare the end state from the complete start composition. Move both fox layers 120 pixels to the right, leaving the background and axis untouched. On this canvas, that is 7.5% of the image width.
Both horizontal displacements are positive and equal because the reflection axis is horizontal. The inverted fox does not travel left. Their vertical distances from the line, sizes, and orientations stay unchanged.
This is a graphic slide, so the fox keeps its seated pose. Asking for walking at the same time would introduce a different task involving changing legs, folds, and contact. Leave that for a separate study.
Check the whole destination before exporting. The nose, tail, and inverted ears need room at the end as well as at the start. Avoid solving a clipped ear by shrinking only one layer; that would break the pairing.
Export two complete images with identical canvas dimensions. The comparison layout below explains the plan, but its labels and side-by-side panels are not the two input images.
Start and planned end: both fox layers shift 120 pixels right on the 1600×900 canvas. The axis, vertical spacing, scale, background, and framing stay fixed. These are externally prepared planning states.
Give the optional transition a narrow brief
If you want to attempt the transition, the HeyDream free image-to-video tool is one option for the generation step. Its current workflow asks for a start image, an end image, a motion or camera description, and an aspect ratio. This pair uses 16:9.
A suitable brief is:
Keep the camera fixed. Slide the upright paper fox and its vertically inverted counterpart gently to the right toward the supplied end positions. Keep the horizontal reflection axis, vertical spacing, scale, seated pose, paper folds, and background unchanged. Preserve the two figures as one matched reflection design.
The words describe the intended constraint. They do not guarantee the generator will preserve it. Layer preparation, inversion, alignment, and annotation happen outside the tool.
Disclosure: I am the founder of HeyDream.
Review the relationship between the endpoints
If you generate a clip later, inspect several intermediate moments as well as the beginning and end. Compare the noses, ear tips, and cream tail folds. Does each upper detail still have one lower counterpart? Do both figures remain equally displaced along the same fixed axis?
Watch for a second tail appearing, one fox changing scale, or the lower figure turning into an independently posed character. Those changes would undermine this particular study even if the movement looked attractive.
Keep the original still pair available. If the transition weakens the relationship, the still composition can carry the process note on its own. The useful result of planning is a clear rule you can inspect—not an obligation to animate every illustration.