Motion control needs a driving clip
Motion control is a still plus a driving clip. The clip is the performance. A prompt is not a substitute.
Motion control is a still plus a driving clip. The still is who. The clip is the performance — timing, weight, the beat of the gesture. A text prompt is not a substitute for the clip. If you only have a photo and a sentence, you are on image-to-video, not motion control.
Search Console still shows this site for "motion control" (35 impressions, average position 70, seven days to 4 September 2026). The existing how-to, Kling V3 motion control guide, is puppeteering mascots. This page is the object. Two files go in. One of them has to move.
Two files, not a better prompt
Motion control transfers the movement in a driving video onto a different subject. The performance stays. The performer changes. Choreography, gesture rhythm, and the beat of a point-at-the-label survive. Wardrobe, lighting, and set come from the character image and whatever prompt the row still accepts.
That is the only mode where you can specify motion exactly. "She spins twice and points at the bottle" is a paragraph with an uncertain outcome. A two-second clip of someone doing it is unambiguous. The model reads pose and timing from the driver and re-renders it wearing the still.
The practical constraint is that the driving clip has to be readable. Full-body framing with clean separation from the background transfers. Crowded frames, extreme close-ups, and heavy occlusion do not, because the pose the model extracts is only as good as what it could see. Body proportions between driver and subject affect fidelity; wildly different builds produce odd limb work. Camera movement in the driving clip may or may not transfer, depending on the model. Do not assume a handheld pan becomes a matching pan on the still.
Facial performance usually does not come across. Most transfer is body-level. Expression and lip movement need a lipsync pass after, if the cut is a talking head. A dance trend on a mascot is motion control. A founder line on a locked still is not.
The driving clip sets the length. You are casting a performance, not editing one. If the driver is six seconds of a spin, the output is that spin, not a 15-second film you described in the prompt box. Trim the driver to the move you want transferred. Do not hand the model a 40-second phone dump and hope it finds the beat.
The Versely rows that actually take a driver
On Versely the published motion-control cards are Kling V3.
Kling Video V3 Pro Motion Control transfers motion from a driving video to a reference image at pro-tier quality. Category: motion-control. Audio is false. 1080p. 17 credits. The file is silent. That is the point. Quoted lines in the prompt do not become a soundtrack.
Kling Video V3 Standard Motion Control is the sibling transfer: same two-file job, audio false, 720p, 13 credits. Use Standard when 720p is the delivery. Use Pro when the listing needs the 1080p plate. Do not "upgrade in the prompt." Change the row.
Neither slug lists a duration enum on the published catalog brief. Do not invent a 5 / 10 / 15 ladder on motion control. Length follows the driver. Kling 3.0 itself shipped in February 2026 with native 4K on other rows (Magic Hour's release tracker, read 4 September 2026); these two motion-control cards are 1080p and 720p as published. The 4K Kling text-to-video row is a different job.
The door is the AI video generator with the motion-control row selected, not a text-to-video box you keep retrying. If the UI asks for a reference image and a driving video, you are in the right object. If it only asks for a prompt, you left.
Audio false means you will add sound later if the cut needs it: a bed, a VO, a lipsync pass on a different row. Prompting "upbeat track, she says the slogan" on a silent transfer wastes the sentence. The mouth will not speak. The file will not contain music because you named a genre.
What this is not
Image-to-video. Image-to-video animates a still you supply. The picture is the opening frame. The prompt describes what happens next. Motion is invented. That is the right tool for a product turn, a slow push-in, a hair-in-the-wind beauty still — generic motion a sentence can specify. It is the wrong tool when the joke is a specific choreography. I2V will give you a dance. Motion control gives you that dance.
Text-to-video. Text-to-video invents every frame from words. No still, no driver. Useful for establishing shots and abstract b-roll. Useless when the character has to be the approved mascot and the move has to match a trending audio's counts.
Reference-to-video. Consistency of identity across clips is a reference problem. Motion control is a performance problem. Same character, many clips, no specific choreography: reference. Same character, this exact bit: motion control.
Lipsync. A driving audio track on a face is a mouth pass. A driving video of a body is motion control. Swap them and you will either get a talking still that does not dance, or a dancing still with a closed mouth. Both can be correct jobs. They are not the same row.
A timeline effect. CapCut-style keyframing, speed ramps, and motion tracking are editorial. Motion control is a generate that re-performs a clip on a new subject. You do not get an editable skeleton. You get a new MP4.
The how-to for brand mascots — film yourself, transfer the bit, keep the timing — lives on the Kling V3 motion control guide. This page stops at the object: still plus driver, silent Kling rows, 17 or 13 credits.
The test before you spend the credits
List what you actually have.
A still and a sentence about motion: image-to-video. Write the camera and the action. Do not open motion control and paste the sentence into an empty driver slot.
A still and a clip of the move: motion control. Check the driver is readable — full body or clear limbs, uncluttered background, trimmed to the beat. Check the still is the character you want wearing that move, not a crop that hides the legs you need to transfer. Pick Pro at 17 credits for 1080p or Standard at 13 for 720p.
A clip of a person who already is the talent, and you only need a new line: lipsync, not motion control. You are not changing the performer.
No still, no driver, a cinematic idea: text-to-video. Motion control will not invent the performer for you.
If the muted result has to hold as a performance — a dance, a signature gesture, a physical joke — the driver is the brief. If the muted result only has to look like the product moving a little, you do not need this object. Save the 17 credits for the week the choreography is the ad.
Kling's other V3 cards will invent picture from text. These two will not. Bring the photo. Bring the clip that already did the move. Then generate.
Generate the transfer from the video generator after you have both files, or sign up and pick the Kling motion-control row that matches the delivery resolution.
FAQ
Can I describe the dance instead of filming a driver?
Not on a motion-control row. The clip is the specification. A prompt is image-to-video or text-to-video, which will invent a move. If the timing has to match a sound or a bit you already performed, film or find the driver.
Does Kling V3 Pro motion control include audio?
No. The published card marks audio false. 1080p, 17 credits, silent. Add a bed, a VO, or a lipsync pass on a different row if the cut needs sound.
Is the driving clip the same as a reference video for identity?
No. A driving clip is the performance to copy. A reference still or reference-to-video pack is who should appear. Motion control uses the still as the performer and the clip as the move. Do not feed a beauty take as a "driver" if you only wanted the face to stay consistent.