Kling 3.0 Standard Motion Control

Kling 3.0 Standard Motion Control (Video-to-Video)

What is Kling 3.0 Standard Motion Control?

Kling 3.0 Standard Motion Control, from Kuaishou's Kling team, is a video-to-video motion-transfer model. Give it two inputs — a single still character image and a 3-30 second reference video — and it produces a new 720p clip in which your character performs the reference's exact movement while keeping its own face, expression and outfit. Because it separates appearance (the image) from action (the video), you can reuse one motion reference across many characters. The standard tier is built for drafts and high-volume iteration before a 1080p pro final render.

Key Features

• •Reference-driven motion transfer: the character copies the driving clip's choreography step for step, instead of guessing from a text prompt.
• •Identity preservation: face, expression and costume stay stable through head turns, fast motion and partial occlusions.
• •Physics-accurate grounding: feet stay anchored and weight shifts look natural, with no sliding.
• •Orientation control: character_orientation follows the reference (up to 30s) or keeps the image's facing (up to 10s).
• •Scene prompt and audio: an optional prompt adds background or camera, and keep_original_sound carries the reference audio.

Best Use Cases

Apply a trending dance, a martial-arts form or an expressive gesture to a custom character, avatar or concept-art figure. Drive brand mascots and digital characters with real performances, and reuse one motion template across a whole cast to build a consistent series. In testing, a still wushu martial artist cleanly reproduced a driving clip's kung-fu kick-and-block sequence at 720p — face, uniform and courtyard setting held throughout — which makes the standard tier a strong fit for social content, previsualization and character animation at a weekly cadence.

Prompt Tips and Output Quality

Motion comes only from the video, so pick a clean, single-person reference with continuous, moderate-speed motion; avoid cuts, multiple people and very fast action. Use a full-body character image with a clearly visible, reasonably large face to protect likeness. The prompt only adds scene, background or camera, so keep it short and concrete. Output length always matches the reference clip.

How to Use Kling 3.0 Standard Motion Control

Kling 3.0 Standard Motion Control animates a still character with the motion of a reference video. You supply two things: the character as an image, and the movement as a short driving clip. The model copies the clip's full-body motion onto your character and renders it at 720p, keeping the character's face, expression and outfit.

Step by step

1. •Upload the character as image_url — a clear, full-body shot with a large, well-lit face (JPG or PNG, up to 10 MB, aspect ratio between 1:2.5 and 2.5:1).
2. •Upload the driving clip as video_url — one continuous single-person performance, 3-30 seconds, MP4 or MOV. The output runs for the length of this clip.
3. •Optionally add a prompt to set the background, scene elements or camera. The motion always comes from the video; the prompt cannot change it.
4. •Choose character_orientation: video follows the reference's facing (clips up to 30s); image keeps the character image's pose (up to 10s).
5. •Set keep_original_sound to carry the reference clip's audio onto the output.

Recommended Settings

• •Character image: single subject, full body, face large and well-lit — this is what protects likeness.
• •Reference video: one person, continuous moderate-speed motion, no cuts or scene changes. Fast or crowded clips reduce reliability.
• •character_orientation: video for full motion transfer; image when you need to lock the character's framing.
• •keep_original_sound: true to retain the clip's sound (a silent reference produces silent output).
• •Iteration: keep references short for quick drafts; the standard tier is built for iteration before a 1080p pro final render.

Tested at these settings, a still martial artist reproduced a driving clip's kung-fu form at 720p with identity and the courtyard setting intact, and with grounded, non-sliding footwork.

Examples to try

Each example ran end to end on the standard tier: a character image plus a 3-30 second reference clip go in, and a 720p clip comes out whose length, motion and timing are the clip's while the character's face and outfit carry over.

• •Reference-driven motion transfer. Prompt: "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 with face, uniform and grounded footwork intact. Tip: use a full-body character image with a large, clear face to protect likeness.
• •A different motion class — dance. Prompt: "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 clip's sweeping arms and turns while her ruffled dress swings and her identity holds. Tip: pick a single-person reference with continuous, moderate-speed motion and no cuts.
• •Identity on a stylized character. Prompt: "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. Tip: character_orientation "video" follows the clip's facing; switch to "image" to lock the figure's pose.

Try other single-person references: "a street dancer running through a popping routine", "a tai chi practitioner flowing through a slow form", "a tennis player serving on a clay court".

Does it keep my character's face?

Yes — identity, expression and outfit are preserved through turns and occlusions.

What resolution does the standard tier produce?

720p; the pro tier renders 1080p for final delivery.

How long can the output be?

Up to 30 seconds following the reference, or 10 seconds keeping the image's facing; length matches the reference.

Can the prompt change the motion?

No — motion comes only from the reference video; the prompt adds scene elements or camera. For reliable results, use a single-person reference with continuous motion and no hard cuts.