Kling 3.0 Motion Control: put a reference video's motion on any character
Kling 3.0 Motion Control transfers a reference video's full-body motion onto a still character, keeping its face, expression and outfit. It ships in two tiers: pro at 1080p for final delivery and standard at 720p for drafts and iteration.
Kling 3.0 Motion Control, from Kuaishou's Kling team, copies the motion of a reference video onto a still character image. You give it who appears (the image) and how they move (a 3-to-30-second driving clip), and it returns your character performing that exact motion, with its own face, expression and outfit intact. Because you show the model the movement instead of describing it, results repeat more reliably than prompt-only image to video. It ships in two tiers: a pro tier at 1080p for final delivery and a standard tier at 720p for drafts and iteration.
Pro tier (1080p)
The pro tier renders the transfer at 1080p, for final delivery and client work. It takes two required inputs: a character image (image_url) and a reference video (video_url). An optional prompt adds scene or camera, and character_orientation chooses whether the body follows the reference's facing or keeps the image's pose. The face, expression and outfit stay stable while the body moves, through head turns and when hands or props briefly cover the face.
ModelKling 3.0 Pro Motion ControlTransfer reference-video motion onto any character in 1080p.The clearest way to see what it does is to watch the reference clip and the result together.
- Reference clip
- Pro output (1080p)
The red-sashed wushu character reproduces the reference's kung-fu form — guard, high kick and lunge punch — with its face, costume and footing held at 1080p.
The flamenco dancer performs on a dark wooden tablao stage under a single overhead spotlight, steady wide shot capturing the full figure.
The dancer copies the reference's sweeping arm turns and stepping footwork while the ruffled red dress flares and her face and costume stay fixed.
The armored warrior-mage performs in a vaulted stone hall lit by blue braziers, steady wide shot capturing the full figure.
The armored warrior-mage turns from profile to face the camera and raises both arms while every engraved plate, glowing rune and the blue cloak stays recognizable.
Use the pro tier from the API
Send the character image and reference clip as URLs, with an optional scene prompt; the response returns the finished 1080p clip.
# pip install "segmind>=1.1.0"
# export SEGMIND_API_KEY="YOUR_API_KEY"
import segmind
result = segmind.run(
"kling-3-motion-control-pro",
image_url="https://segmind-resources.s3.amazonaws.com/input/kling-3-mc-pro-character-9f30d367.jpg",
video_url="https://segmind-resources.s3.amazonaws.com/input/kling-3-mc-pro-driving-79a412a6.mp4",
prompt="The martial artist performs the form on a grey flagstone temple courtyard, steady wide shot capturing the full figure.",
character_orientation="video",
keep_original_sound=True,
)
print(result.get("output"))# 1. Submit the request; the reply carries its request_id
REQUEST_ID=$(curl -s -X POST "https://api.segmind.com/v2/kling-3-motion-control-pro" \
-H "x-api-key: YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"image_url": "https://segmind-resources.s3.amazonaws.com/input/kling-3-mc-pro-character-9f30d367.jpg",
"video_url": "https://segmind-resources.s3.amazonaws.com/input/kling-3-mc-pro-driving-79a412a6.mp4",
"prompt": "The martial artist performs the form on a grey flagstone temple courtyard, steady wide shot capturing the full figure.",
"character_orientation": "video",
"keep_original_sound": true
}' | jq -r .request_id)
# 2. Check its status until it is COMPLETED
curl -s "https://api.segmind.com/v2/requests/$REQUEST_ID/status" \
-H "x-api-key: YOUR_API_KEY"
# 3. Fetch the result
curl -s "https://api.segmind.com/v2/requests/$REQUEST_ID" \
-H "x-api-key: YOUR_API_KEY"Standard tier (720p)
The standard tier runs the same workflow at 720p, built for drafts and high-volume iteration before a pro final render. The inputs and settings are identical: a character image, a reference clip, an optional scene prompt, orientation and audio. Rough out a motion on standard, then switch to pro for the 1080p delivery.
ModelKling 3.0 Standard Motion ControlTransfer reference-video motion onto a character image at 720p.The martial artist performs the form on a grey flagstone temple courtyard, steady wide shot capturing the full figure.
The still wushu martial artist reproduces the clip's kung-fu kick-and-block form at 720p, with face, uniform and grounded footwork intact.
The armored warrior-mage performs in a vaulted stone hall lit by blue braziers, steady wide shot capturing the full figure.
The ornate silver armor, glowing accents and blue cape stay stable through a profile turn and an overhead arm-raise.
Use the standard tier from the API
The call is the same as the pro tier, pointed at the standard slug; the response returns a 720p clip.
# pip install "segmind>=1.1.0"
# export SEGMIND_API_KEY="YOUR_API_KEY"
import segmind
result = segmind.run(
"kling-3-motion-control-standard",
image_url="https://segmind-resources.s3.amazonaws.com/input/kling-3-mc-std-character-4c224715.jpg",
video_url="https://segmind-resources.s3.amazonaws.com/input/kling-3-mc-std-driving-17af71bf.mp4",
prompt="The martial artist performs the form on a grey flagstone temple courtyard, steady wide shot capturing the full figure.",
character_orientation="video",
keep_original_sound=True,
)
print(result.get("output"))# 1. Submit the request; the reply carries its request_id
REQUEST_ID=$(curl -s -X POST "https://api.segmind.com/v2/kling-3-motion-control-standard" \
-H "x-api-key: YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"image_url": "https://segmind-resources.s3.amazonaws.com/input/kling-3-mc-std-character-4c224715.jpg",
"video_url": "https://segmind-resources.s3.amazonaws.com/input/kling-3-mc-std-driving-17af71bf.mp4",
"prompt": "The martial artist performs the form on a grey flagstone temple courtyard, steady wide shot capturing the full figure.",
"character_orientation": "video",
"keep_original_sound": true
}' | jq -r .request_id)
# 2. Check its status until it is COMPLETED
curl -s "https://api.segmind.com/v2/requests/$REQUEST_ID/status" \
-H "x-api-key: YOUR_API_KEY"
# 3. Fetch the result
curl -s "https://api.segmind.com/v2/requests/$REQUEST_ID" \
-H "x-api-key: YOUR_API_KEY"Side by side
Running the same character image and reference clip on each tier isolates the one thing that changes between them: resolution.
- Pro (1080p)
- Standard (720p)
The motion, timing and identity match across both; the pro clip carries more fine detail at 1080p than the standard clip at 720p.
Shared settings
| Setting | Value | Members | When to use it |
|---|---|---|---|
image_url | Character image, JPG/PNG up to 10 MB, aspect 1:2.5 to 2.5:1 | Both | Required. A clear full-body shot with a large, visible face protects likeness |
video_url | Reference clip, MP4/MOV, 3 to 30 s | Both | Required. The motion source; the output runs this clip's length |
prompt | Up to 2500 characters, scene or camera only | Both | Optional. Adds background or camera, never the motion |
character_orientation | video (default) or image | Both | video follows the reference (up to 30 s); image keeps the image's pose (up to 10 s) |
keep_original_sound | true (default) or false | Both | Keep the reference clip's audio, or turn it off for a clip you will score yourself |
| Output resolution | 1080p / 720p | Pro / Standard | Pro for final delivery; standard for drafts and high-volume iteration |
Good to know
- Motion comes only from the reference video. This is motion transfer, not text to motion — the prompt adds scene, background or camera, but cannot invent movement.
- Use a clean, single-subject reference. Multiple people, hard cuts, scene changes or very fast motion reduce reliability; in a multi-person clip, the character filling the largest area drives the motion.
- Give the character image a clearly visible, reasonably large face. A small or soft face is the main cause of likeness drift.
- Output length always matches the reference clip: up to 30 seconds following the reference (
character_orientation=video), or up to 10 seconds keeping the image's pose (character_orientation=image). keep_original_soundcarries the reference clip's audio onto the output; a silent reference produces a silent clip.
Resources
- Try Kling 3.0 Pro Motion Control on Segmind · API reference
- Try Kling 3.0 Standard Motion Control on Segmind · API reference
- Kling Motion Control User Guide — How to drive character movement, facial consistency and cinematic detail with Motion Control; choosing a reference clip and the orientation and scene options
- Kling AI Motion Control — Character Animation & Motion Capture — The Motion Control feature overview: the 3-step workflow and why it replicates movement and expression from a reference
- Ultimate Guide to Kling Motion Control (video tutorial) — A step-by-step walkthrough for making a motion-transfer video, including reference-clip selection, scene prompting and orientation tips
- #kling-3-motion-control
- #kling
- #model-launch
- #video-to-video
- #motion-transfer
