How to Film a Motion Reference Video for Kling Motion Control on Your Phone

In Kling Motion Control, your video clip supplies the body movement, timing, facial expressions, and physics, which means most fixable animation errors start in the reference footage before you ever touch the generator. When you record that clip on a phone at home, treating the recording like a tiny film shoot saves you from blurry limbs, cropped feet, and stiff results.

Pick a Move That Transfers Cleanly

Kling introduced Motion Control in Kling VIDEO 2.6, an update that improved complex moves such as martial arts and dance. Even with those improvements, real human motion recorded at a moderate pace transfers more accurately than frantic or stylized actions. According to Kling's guide, very fast or complex movement may produce shorter, less accurate output.

Before recording, consider the proportions of your character image. A human illustration or 3D humanoid render can follow a full-body wave, a slow spin, or a boxing stance easily. A short, round mascot copying high kicks often looks odd because its body does not match yours. Pick motion that suits the character's physique, and practice it at a steady pace so your hands and feet stay easy to track.

Set Up Your Room and Prop Up the Phone

You do not need a studio or a helper to shoot a clean motion reference video. You only need a plain wall, even light, and a still background. Turn on the room lights or place a lamp in front of you so your face and body are lit evenly without deep shadows. Clear away clutter, spinning fans, and pets that might wander behind you.

Camera movement quickly confuses Kling motion transfer. Platform docs from Vercel and others specifically advise against camera movement in the reference clip. Even a friend holding the phone by hand introduces subtle swaying. Instead, set the phone on a tripod, or prop it upright against a stack of books on a table or shelf so the camera stays completely still.

Match Your Framing to the Character Image

In Kling's own guide, mismatched framing between the character image and the reference video is one of the most common mistakes. Pairing a tight face close-up with a full-body dance forces the model to invent missing limbs or ignore most of the choreography.

Decide on your framing before you step back from the phone, and make sure it matches the character image.

• Pair a full-body character image with a full-body reference video where your head, hands, and feet stay visible the entire time.
• Pair a waist-up character image with a waist-up video where your arms and hands never drop out of view.
• Leave space above your head and below your feet so reaches or small steps never touch the frame edge.

Your character image needs that same room for movement. Choose one clear character (real person, illustration, 3D render, or mascot) with head and body fully visible, nothing cropped or covered, and the short edge at least 300 pixels. Keep responsible use in mind as well. Do not animate a real person without their permission, and label realistic results of real people as AI-generated. Original characters avoid the problem altogether.

Dress for Contrast and Perform in One Take

Wear fitted clothes that contrast with the wall behind you. Against a light wall, wear a dark shirt and pants. Long skirts, capes, and big sleeves can hide limbs and joints, making it harder for the tool to read the skeleton of your performance.

Keep the recording to one person in one continuous take with no cuts, and stay within the length limit shown in the tool. Make sure nobody walks through the frame.

Start in a pose close to the character image's pose and hold it briefly before moving. If your character has both arms relaxed at their sides, start the same way before raising a hand. A starting pose close to the image's pose transfers more cleanly than one that begins in a completely different position.

Run a One-Minute Check on Your Phone

Before uploading the clip, play it back on your phone and verify five details across the take.

• Hands, head, and feet stay in frame the whole time.
• The wall edges remain completely still, showing the camera did not move.
• Only one person appears, with nobody walking through the background.
• The clip is one continuous shot with no cuts.
• The speed looks moderate, without motion blur streaking your hands.

Trim the dead seconds at the start and end (where you walked from the phone and walked back to press stop) in your phone's editor so the video opens on your starting pose.

Check Facing Direction and Scene Prompts Before Uploading

When you bring your clip to Kling5 AI, an independent third-party site that is not Kling's official site and is not affiliated with Kuaishou or the official Kling team, you will find Motion Control as a mode of the Kling 3.0 generator next to text-to-video and image-to-video. (Kling 5 has not been announced or released.)

Kling's official app and API offer two orientation modes (follow the video's facing, or keep the image's facing), whereas Kling5 AI has no orientation switch. On the site, the character's facing always follows the uploaded character image, so pick an image that already faces the way you want.

Once your video and character image are loaded, use the text prompt only to set the scene (background, lighting, and style) rather than redescribing the moves. Readers who want the complete step-by-step on the site can read the full Kling Motion Control walkthrough on Kling5 AI.

If you want to build longer scenes around your character's performance, the Kling 4.0 generator on the site offers reference-to-video and clips up to 30 seconds instead of Motion Control; our guide on how to brief a longer Kling 4.0 shot like a director explains how to plan those shots. For the performance clip itself, five minutes spent propping your phone against a stack of books and checking your framing against a plain wall will do more for the final animation than any prompt tweak afterward.