Use the 19-second Apartment shadowboxing template
Watch Apartment shadowboxing in the template list and click Use. Sign in with a verified account and use an existing balance or buy the credits needed for the scene. Select a clear photo you have permission to use, confirm your rights, then click Upload. Wait for both the photo and template reference checks to pass, and approve the photo. Check the HD 720p quote before generating. No newly generated portrait is required for this template route. Choose the shadowboxing scene.
Use a photo that makes the subject easy to identify, with an unobstructed face and no competing people. Motion transfer uses the reference to guide the performance, but a photo and a reference clip cannot guarantee exact facial similarity, anatomy or every punch. Review the finished hands, face and movement before downloading and sharing.
Frame an original scene so the jabs can be seen
For a new scene, compose the still image around the movement. Show the torso, both hands and enough empty space in front of the fists. A close face crop leaves the model little room to represent the punch. A full-body shot includes footwork but makes the face smaller, so choose the framing according to which part of the joke matters.
Copy-ready portrait prompt: A fictional adult man with a blunt blond bob, curled moustache and a brown houndstooth suit stands on a quiet Paris street with pale stone buildings. Deadpan expression, elbows bent, both fists visible in a relaxed guard. Three-quarter body view, open space around the arms, natural afternoon light, realistic hands and proportions, vertical composition.
Treat Paris as an imagined setting for your own creation. This prompt is an original example; it is not a recovered private prompt or proof of where a source video was filmed.
Start with one jab and a return to guard
Copy-ready motion prompt: The man holds a completely serious expression. He throws one short left jab into empty space, draws the fist back to guard, then makes a small shift of weight and pauses. Preserve the face, blond bob, curled moustache, suit and street. Static eye-level camera, one continuous shot, no cuts or other people.
Start with a short duration supported by your selected renderer. If the first result is readable, add a second action in a later variation. A sequence of several punches, a spin, a zoom and a changing location asks the model to solve several problems at once, making it harder to identify which instruction caused an awkward result.
Use a licensed reference when timing matters
An image-to-video prompt invents the timing. A motion-control workflow takes an image plus a reference video to guide the movement instead. Higgsfield’s official Kling 2.6 Motion Control documentation describes that input pair. Select a clear, licensed reference with the movement you actually want rather than expecting text alone to reproduce someone’s exact routine. Read the official motion-control input requirements.
On this site, Create your own → Use my own motion clip accepts a 3–30 second reference. A new character portrait costs $1.00 separately. Its HD motion video costs $9.90 for up to 19 seconds; longer references use the rounded duration quote shown before generation. Record a restrained shadowboxing performance yourself if you need a reference you can directly control.
Inspect the action, then decide whether it needs sound
Watch at normal speed first: the action should read as a punch and return, with a visible pause. Then inspect the hands, jacket edges and face more closely. Check that a fist does not fuse with the torso and that the background does not jump during the jab. These are practical review points, not a claim that a prompt can prevent every artifact.
Template videos and the site’s five-second idea videos are silent. You can add music you are allowed to use in an editor after downloading. A soundtrack can accent the pause, but it should not hide a movement that fails to make sense without audio. See completed site examples and their provenance. Adapt the portrait and motion prompt separately.