Kling motion control for an AI persona: face, cost, rules

Kling motion control moves your AI persona?AI persona An AI persona is an invented character. Her images, videos and captions are generated with AI instead of photographed, but she is presented and followed like any other account. Open the glossary like the person in a video you upload. Use Kling 3.0, set the character orientation to “match video”, and bind a face element made from your persona’s own images. Kling says this keeps the face more stable. A 5-second clip costs about $0.63 to $0.84, and you should switch the original sound off.

What Kling motion control does with your persona image

Kling motion control takes two things: one image of your character and one video of a person moving. The result shows your character doing that movement, with the same facial expressions.

“Motion Control enables precise control of a character’s movements and facial expressions based on a reference image.” Kling VIDEO 3.0 Motion Control is dated 5 March 2026. It “enhances facial consistency across scenarios, ensuring stable facial features and smooth expressions even in complex, multi-angle, long-duration motions.”

Kling Quickstart — Motion Control user guide (opens in a new tab)checked 5 Oct 2026

For a persona this solves the biggest problem of image-to-video: made-up movement. A real person does the dance, the walk or the gesture. The model only has to put your character into it. The best AI video models for an AI influencer compares the other modes.

The face still comes from your image. A weak image gives you a weak face that moves very well. And the driving video has to be one you are allowed to use, as the rules section explains.

Kling 4.0 has been live in the Kling app since 15 October 2026. Its Omni Reference can take a motion reference video together with a reference image of your character. The dedicated motion control tool is still the 3.0 version described here. Kling’s motion control guide only covers VIDEO 3.0. Kling’s API offers 4.0 in early access, but its motion control rows are still for 3.0 and 2.6. So every price and API setting below is for 3.0.

If you try 4.0 with a motion video through the API, Kling counts the video as input. A 10-second clip with a 10-second motion video is billed as 20 seconds: $1.88 at 720p. Kling 3.0 motion control costs $1.26 for the same clip on Kling’s API.

“The all-new Kling 4.0 officially became available to everyone on October 15.” Video references can “guide character performances, actions, camera movements, and narrative pacing.”

Kling AI Release Notes — Meet All-New Kling 4.0 (opens in a new tab)checked 21 Oct 2026

“【New Release】Kling 4.0 API Is Now Available for Early Access”. Kling 4.0, “With Input Video”, “Input Duration ≤ Output Duration”, “Total duration = Output duration × 2”: “$0.094/s” at 720P, “$0.141/s” at 1080P. Billing details: “Input video duration includes reference video duration and video character elements duration.” Kling 3.0, “Motion Control”, “Total duration = Output duration”: “$0.126/s” at 720P, “$0.168/s” at 1080P.

Kling AI API docs — Video pricing (opens in a new tab)checked 21 Oct 2026

The whole routine for one persona clip has five steps. Each section below covers one of them.

1. Film your own taketripod, one person, 5 to 10 s
2. Make four stillsfront, left, right, smiling
3. Bind a face element1 main image and up to 3 more
4. Match video, sound offorientation and audio
5. Test short, then go long5 s first, up to 30 s

Schematicthe shape, not measured data

Kling motion control character orientation: match video or match image

Choose “match video” for a persona. It allows clips up to 30 seconds and it is the only setting where you can bind a face. Choose “match image” only when you need a camera movement from your prompt.

“Character Orientation Matches Image”: “Camera movements and other elements can be customised through prompts.” “Element Binding is supported only when the character’s orientation matches the video orientation.”

Kling Quickstart — Motion Control user guide (opens in a new tab)checked 5 Oct 2026

“image: Matches the character orientation in the reference image. Reference video duration must not exceed 10s.” “video: Matches the character orientation in the reference video. Reference video duration must not exceed 30s.” “When using an Element, the generated video always follows the character orientation from the reference video.”

Kling AI API docs — Motion Control 3.0 (opens in a new tab)checked 5 Oct 2026

	Match video	Match image
Movement and expressions	From the video	From the video
Which way she faces	From the video	From your image
Camera movement	From the video	From your prompt
Longest driving video	30 seconds	10 seconds
Face binding (Element)	Yes	No
Use it for	Dances, walks, turns, anything long	A fixed pose with a slow zoom or tilt

If you pick match image to control the camera, you lose the face lock. For a persona, the face matters more.

Element Binding: how to keep your persona’s face in Kling motion control

Bind a face element made from two to four images of your persona: a front view first, then the angles and expressions the clip needs. Kling uses the element only for the face, not for hair or clothes.

“The Motion Control Element Library only uses facial information for reference. It does not include clothing, hairstyle, makeup, or props. Therefore, we recommend uploading clear facial close-ups to ensure sufficient facial data.” For head turns, upload “A front-facing view, side views (left and/or right)”.

Kling Quickstart — Motion Control user guide (opens in a new tab)checked 5 Oct 2026

“Each element must contain at least 2 reference images (1 main reference image + 1 additional reference image) and can include up to 4 reference images (1 main reference image + 3 supplementary reference images).”

Kling Quickstart — Element Library user guide (opens in a new tab)checked 5 Oct 2026

How to set it up in the Kling web app:

1. Make the stills first. Generate a front close-up, a left and a right three-quarter view and one smiling close-up of your persona. ChatGPT AI influencer photos shows how to get the same face across such a set.
2. Upload your driving video and your character image. The character image shows the whole body or upper body, matching the framing of the video.
3. Click “Bind Facial Element to Enhance Facial Consistency” below the character image. Create a new element with the front close-up as the main image and up to three more.
4. Keep “Character Orientation Matches Video”. Binding switches off with image orientation.
5. Name the element after your persona and reuse it for every clip, so every video is measured against the same face.

Two limits from Kling’s own guide. Only one element works per clip, and Kling picks the person who takes up most of the frame. “If the elements occupy similar portions of the frame, no element will be selected.” So keep one person in your driving video.

Kling also lets you record a video of yourself to create an element. For an invented persona, use her generated images only.

What Kling motion control costs per clip: app, API, WaveSpeed, fal, Higgsfield

On 5 October 2026 Kling 3.0 motion control cost the same per second through Kling’s own API, WaveSpeed and fal: $0.126 at the lower tier and $0.168 at the higher tier. A 5-second test clip costs $0.63 or $0.84. The Kling app counts in credits, and Higgsfield shows its price only before you generate.

Where	Price	5 s clip	10 s clip	30 s clip	Face element	Original sound
Kling app, Standard	9 credits/s	45 credits	90 credits	270 credits	Yes	Not stated in the guide
Kling app, Professional	12 credits/s	60 credits	120 credits	360 credits	Yes	Not stated in the guide
Kling API, 720p	$0.126/s	$0.63	$1.26	$3.78	Yes, one	On by default
Kling API, 1080p	$0.168/s	$0.84	$1.68	$5.04	Yes, one	On by default
WaveSpeed Std / Pro	$0.63 / $0.84 per 5 s	$0.63 / $0.84	$1.26 / $1.68	$3.78 / $5.04	Listed, not explained	On by default
fal Standard / Pro	$0.126 / $0.168 per s	$0.63 / $0.84	$1.26 / $1.68	$3.78 / $5.04	Yes, one	On by default
Higgsfield	Credits	Shown on the Generate button			Not stated	Not stated

Prices are from 5 October 2026. The clip prices are the per-second rate times the length. What generation costs lists the other video models with their dates.

What the table shows:

• The resellers add nothing on top. WaveSpeed’s Std price equals Kling’s 720p API price, and its Pro price equals the 1080p price.
• The face element is documented best on Kling’s API and fal. Both say one element per clip, which matches Kling’s own guide. WaveSpeed lists an element_list field with up to three items, but its page does not say how you create an element for it. Run one test clip there before you build on it. The WaveSpeed review covers its other strengths.
• A failed clip still costs money. Kling’s guide says that when the output comes out shorter because the movement was too fast, “the consumed Credits are non-refundable”. Test at 3 to 5 seconds first.

If you already pay for Higgsfield, you can use motion control there too. Higgsfield’s help pages do not list a credit price for it, so read the Generate button and divide by the seconds.

“Every generation costs a certain number of credits depending on the model, resolution, and duration, and the exact cost is shown on the Generate button before you confirm.”

Higgsfield Help — How credits work (opens in a new tab)checked 5 Oct 2026

The prices in the table come from each provider’s own page:

Kling 3.0, Motion Control: “0.9 Units ($0.126) /s” at 720P and “1.2 Units ($0.168) /s” at 1080P.

Kling AI API docs — Video pricing (opens in a new tab)checked 5 Oct 2026

Kling VIDEO 3.0 Motion Control: Professional “12 Credits/s”, Standard “9 Credits/s”.

Kling Quickstart — Motion Control user guide (opens in a new tab)checked 5 Oct 2026

“$0.63 per 5 seconds”. “Minimum charge: 3 seconds”. “Key inputs: prompt, image, video, negative_prompt, character_orientation, element_list.” keep_original_sound “(default: enabled)”.

WaveSpeedAI — Kling V3.0 Std Motion Control (opens in a new tab)checked 5 Oct 2026

“$0.84 per 5 seconds”, with the same parameters as the Std version.

WaveSpeedAI — Kling V3.0 Pro Motion Control (opens in a new tab)checked 5 Oct 2026

element_list: “Element reference list”, “0 ~ 3 items”. keep_original_sound default true. character_orientation default “video”.

WaveSpeedAI API docs — Kling V3.0 Std Motion Control (opens in a new tab)checked 5 Oct 2026

“Price: $0.168 per seconds”. elements: “Optional element for facial consistency binding.” “Only 1 element is supported.” keep_original_sound “Default: true”.

fal — Kling Video v3 Pro Motion Control (opens in a new tab)checked 5 Oct 2026

“Price: $0.126 per seconds”.

fal — Kling Video v3 Standard Motion Control (opens in a new tab)checked 5 Oct 2026

How to film a driving video for Kling motion control

Film it yourself on your phone: one person, one continuous take, no camera movement, steady speed, framed like your character image.

Video: “The character’s upper body or full body must be clearly visible, including all limbs and the head. Avoid occlusion.” “Avoid cuts, camera movement, etc.; otherwise, the video may be trimmed.” Formats “.mp4 / .mov”, at most 100 MB, sides between 340 px and 3850 px. Image: “Avoid driving a half-body character with full-body motion.”

Kling AI API docs — Motion Control 3.0 (opens in a new tab)checked 5 Oct 2026

1. Record yourself, or someone who agreed. The output shows your persona’s face with your movement.
2. Put the phone on a tripod or a shelf. Camera movement in the driving video cuts the clip short.
3. Match the framing. Full-body image, full-body video. Half-body image, half-body video.
4. Leave room around the body. For big movements, the character image needs space for her arms and legs to move into.
5. Move at a steady, medium speed. Fast spins and jumps make the model use only part of the take. You pay for that part anyway.
6. Keep your face visible and turned to the camera most of the time. The face element works best with angles you uploaded.
7. Film 5 to 10 seconds first. Go to 30 seconds only once a short version holds her face.

The character image follows the same logic: whole body and head visible, nothing in front of the face.

Kling motion control API: one call and three settings to check

Kling’s own API takes the image, the video and an optional element in one request. Before you send it, check three settings: orientation, resolution and audio.

audio: “original: The generated video retains the original sound of the reference video. off: The generated video has no audio.” Default “original”. resolution default “720p”. Generated results “will be cleared after 30 days. Please make sure to save them promptly.”

Kling AI API docs — Motion Control 3.0 (opens in a new tab)checked 5 Oct 2026

Kling asks you to keep the API key in an environment variable. Here is the call with sound off, in the format of Kling’s own example:

curl --location 'https://api-singapore.klingai.com/motion-control/kling-3.0' \
  --header "Authorization: Bearer $KLING_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
    "contents": [
      { "type": "prompt", "text": "She wears a grey hoodie and jeans. Plain white studio wall behind her." },
      { "type": "image", "url": "https://example.com/persona-full-body.png" },
      { "type": "video", "url": "https://example.com/my-own-take.mp4" }
    ],
    "settings": {
      "character_orientation": "video",
      "resolution": "720p",
      "audio": "off"
    }
  }'

What each part does:

• $KLING_API_KEY reads your key from the environment, so it never sits in a script you share.
• The prompt describes her clothes and the background. The face element carries only her face, so the rest has to come from the image and the prompt.
• The image and the video are public URLs to your persona image and your own driving take.
• "character_orientation": "video" keeps the face binding possible and allows takes up to 30 seconds.
• "audio": "off" stops the sound of your take from ending up in the clip.

The answer contains a task ID. Poll the task until its status is succeeded, then download the video at once. To bind a face, add one element entry with the element ID from Kling’s Element API.

On WaveSpeed and fal the same switch is called keep_original_sound. Set it to false.

Kling motion control problems and how to fix them

Most failed clips have one of six causes. Pick what you see, and the matching cause and fix are highlighted.

What went wrong with your clip?

Shorter than my videoNo face element optionThe element is ignoredFace changes as she turnsSomeone else's musicLimbs warped or cut off

• Motion too fast, or a cutKling only uses the part it can follow, and the credits are not refunded. Film one slow, steady take with a still camera.
• Orientation is set to imageElement Binding only works with match video. Switch the orientation back.
• Two people in the frameKling binds the person who fills most of the frame. If two people are the same size, it binds nobody. Film one person.
• No side views in the elementAdd left and right three-quarter close-ups as the extra element images.
• Original sound is onSet audio to off on Kling's API, or keep_original_sound to false on WaveSpeed and fal.
• Framing does not matchUse a full-body image with a full-body take, and leave room around her for the movement.

Schematicthe shape, not measured data

If none of these fits and her face still drifts, the cause is often the still you started from. Compare the first frame with your reference images before you change anything in Kling.

Wan 2.2 Animate or Kling motion control

Wan 2.2 Animate does the same job for less money per second, at lower resolution and without a face element. Use it to test a driving video cheaply. Use Kling when the face has to hold.

“480p 0.20 USD per 5 s → 0.04 USD per second”. “720p Double 480p → 0.40 USD per 5 s → 0.08 USD per second”. Billed between 3 and 120 seconds. Modes: “animate” and “replace”.

WaveSpeedAI — Wan 2.2 Animate (opens in a new tab)checked 21 Oct 2026

	Wan 2.2 Animate (WaveSpeed)	Kling 3.0 motion control
Price per second	$0.04 (480p), $0.08 (720p)	$0.126 (720p), $0.168 (1080p)
Highest resolution	720p	1080p
Longest clip	120 s billed at most	30 s
Face lock	No face element (a seed?Seed A seed is the number that sets the random starting point of a generation. The same prompt with the same seed produces the same image. Open the glossary setting for repeatable runs)	One face element
Sound	Not listed as an option	Original sound on by default

A good routine: test each new driving video in Wan at 480p for a few cents. If the movement transfers cleanly, render the final clip in Kling with the face element.

Leave Wan’s “replace” mode alone: it puts your character into someone else’s existing shot.

Kling motion control rules: sound, the driving video, commercial use and labels

Four rules matter. You must own or have permission for the image and the video. Commercial use depends on your Kling plan. Kling marks every output. And you label the clip as AI where you post it.

Last updated 21 April 2026. 4.3: you warrant that you hold the rights to the Input “or that you have obtained legal authorization from the relevant owner”, including “personality rights”. 4.6: “without our written permission, you may not use, reproduce, distribute, and create derivative works of, and make modifications to, the Output for any commercial purposes.” 4.7.1: Kling gets a “non-exclusive, royalty-free right and license” to use your Input and Output.

Kling AI — Terms of Service (opens in a new tab)checked 5 Oct 2026

Release date 21 April 2026. Membership benefits: “KLING AI members’ use of the Output for commercial purposes is not restricted.” The exception is products that compete with Kling.

Kling AI — Terms of Paid Service (opens in a new tab)checked 5 Oct 2026

“An imperceptible watermark and metadata are embedded in images and videos”. “Aligned with open content-provenance standards (e.g. C2PA)”.

Kling AI — AI Content Detection (opens in a new tab)checked 5 Oct 2026

What that means for your persona:

• The driving video. A clip from someone else’s TikTok is not your input. Film your own.
• The sound. With the default setting, the sound of the driving video ends up in your post. Switch it off and add music from the platform’s library.
• Commercial use. On a paid Kling plan, the paid terms allow commercial use. On a free plan, the general terms ask for written permission. Through the API or a reseller, read that provider’s terms as well. This is not legal advice.
• Kling’s licence. Kling may use what you upload and generate. Do not upload anything you would not want reused.
• The label. Kling’s invisible watermark does not replace a visible AI label?AI label An AI label is a platform’s own marker that says an account or a post is AI-generated. Open the glossary. The label matrix shows what each platform asks for, and don’t strip your watermarks explains why you keep the metadata in the file.

Start with one clip today. Film yourself doing a 5-second wave on a tripod, upload your persona’s full-body image, bind a face element from her four best close-ups, keep “match video” and switch the sound off. If her face holds next to your reference images, film the longer take. If it does not, fix the stills first with the character consistency guide.