Kling Motion Control: how to use it, the input rules and what a clip costs

This guide covers what Kling Motion Control does, the input rules published by Kling and by API partners, and what a clip costs in Kling credits and on a pay-per-second API.

What is Kling Motion Control?

Kling Motion Control is a feature that takes the movement and facial expressions from a reference video and applies them to a character in a still image.

Kling's feature page lists two versions, 2.6 and 3.0. It says 3.0 improves facial consistency over 2.6 across complex motion, multiple angles and longer sequences. It supports human, anime and 3D characters if the style stays consistent. It supports one primary subject. If several faces appear, the system prioritises the dominant one.

On 3.0 you can also bind a facial element (a set of images or a short video of the face) to lock facial consistency. Kling's user guide says binding works only when the character's orientation matches the video orientation.

Two inputs are required: a character image and a motion reference video. A text prompt is optional. In the ComfyUI integration it is used to change the environment, with a limit of 2,500 characters.

How do you use Kling Motion Control?

You upload a character image and a reference video, pick an orientation and quality, then generate.

1. Pick the reference video. Choose a clip with one clear subject, a simple background and a clean silhouette. Keep it between 3 and 30 seconds and a single continuous shot with the character visible throughout. Kling says cuts, shot changes and camera moves can truncate the output.
2. Pick the character image. Use a clear, front-facing image with strong facial detail and visible limbs.
3. Match the framing. Use a close-up reference with a portrait image, and a full-body reference with a full-body image. Higgsfield's guide says a mismatch can make the face shake or warp.
4. Choose the orientation. Video orientation follows the reference video's movement, camera and facing direction. Image orientation keeps the character facing the way the still image does.
5. Choose Standard or Pro. Standard renders at 720p and Pro at 1080p, per the ComfyUI docs.
6. Optional: add a prompt. Use it to describe the setting, not the motion. The motion comes from the video.
7. Generate, then check hands and face before you spend credits on a longer clip.

What are the Kling Motion Control input limits?

The published limits differ by source, so use the stricter number.

Kling's user guide confirms the 3-30 second range and a 340 px short edge and 3850 px long edge for the reference video. The rest come from the ComfyUI documentation for the Kling partner node. Use 720 px or larger for the image.

What makes Kling Motion Control fail?

Mismatches between the image and the video cause most problems that the sources describe.

• Hidden hands. Hands in pockets or behind the body in the image can lead to artifacts (ComfyUI tutorial, Higgsfield).
• Species or pose mismatch. Kling says a human image with animal motion can degrade the result. Keep subject type and pose aligned.
• Framing mismatch. A close-up reference on a full-body image, or the reverse, can warp the face.
• Back views on flat cartoons. The ComfyUI tutorial suggests 3D-style or realistic characters when the reference rotates, because flat 2D art may struggle.
• No room to move. Leave empty space around the subject so big arm movements do not clip the frame.
• Multiple people. Only one primary subject is supported.
• Fast or complex motion. Kling warns the output can come back shorter than your upload, and the credits used are not refunded. The model needs at least 3 seconds of usable continuous motion.

Is Kling Motion Control free?

No. Motion Control is billed in credits per second of output (see is Kling AI free for what the free plan covers), and none of the sources documents a free Motion Control allowance. Pages that promise "free" or "unlimited" Motion Control state no caps or plan conditions.

The credit amount your account shows before you press generate is the real cost. The table below has the published rates.

How much does a Kling Motion Control clip cost?

In Kling's app, Motion Control is billed per second of output, rounded to the nearest whole second: 5 credits/s for 2.6 Standard, 8 for 2.6 Professional, 9 for 3.0 Standard and 12 for 3.0 Professional. A 10-second 3.0 Professional clip is 120 credits.

On the fal.ai API, a 10-second clip costs $0.70 with 2.6 Standard, $1.26 with 3.0 Standard and $1.68 with 3.0 Pro, as of 2026-10-26.

fal.ai bills on generated output duration, per second, with no minimums or subscriptions. Rates are $0.07/s, $0.126/s and $0.168/s.

Retries change the real price. If you keep one of every three takes, a 10-second 3.0 Pro clip costs $1.68 x 3 = $5.04 per keeper. Plan on about three tries per keeper.

Runway's API price page lists Kling 3.0 and Kling 4 models but no Kling motion control model, so do not use its Kling rates for this feature. For what credits cost in money, see Kling AI pricing.

Which version should you use: Kling 2.6 or 3.0?

Kling 3.0 is the choice when the face must stay consistent, and 2.6 is the cheaper choice for rough drafts.

Kling says 3.0 improves facial consistency across complex motion, multiple angles and longer sequences. On fal.ai, 3.0 Standard costs 80% more per second than 2.6 Standard ($0.126 vs $0.07). The efficient workflow, using the 2.6 vs 3.0 comparison, is to test the framing on 2.6, then render the final on 3.0.

Pre-flight checklist

Run through this before you generate.

• Reference video is 3-30 s (3-10 s if you use image orientation).
• One person, clean background, clear silhouette.
• Image shows head, shoulders and torso, with both hands visible.
• Image is at least 720 px and within a 1:2.5 to 2.5:1 aspect ratio.
• Shot scale matches: close-up with close-up, full body with full body.
• Image and video face the same direction if you use image orientation.
• Character type matches the mover (human to human, not human to animal).
• You own, or have permission to use, the video and the face in the image.
• You have budgeted for about three tries per keeper.

What about faces, rights and platform rules?

Get permission from anyone whose likeness you animate, secure the rights to your reference footage, and use characters you own or have licensed.

Motion Control is often used for dance and meme formats. If the reference video is someone else's dance, you are copying their performance. If the image is a real person, a friend or a celebrity, Kling's platform terms and local likeness law apply, so check both before posting commercial work. For formats built on this idea, see how AI dog dance videos are made and AI fruit love island videos. More are in the Kling AI guides and the Learn hub.