Character, look, and background source. Use a clear character image with visible proportions. JPG, PNG, or WebP up to 10 MB.
Motion source. Use a realistic video with the head and upper body or full body visible. Up to 10 seconds for image angle mode, up to 30 seconds for video angle mode.
Optional guidance for the generated motion video.
Keep the sound from the reference video when available.
Sign in to spend tokens
Reference Image is required
Your generated video will appear here
Kling Motion Control 3 transfers realistic motion from a reference video onto a character image — your character, look, and background move exactly the way the person in the source clip moves. Prizmad ships it as a built-in AI Studio tool, no separate Kling account needed.
Kling Motion Control 3 is a motion-transfer video tool: you upload one character image and one reference video, and the model generates a new clip where your character performs the motion from the video. The image supplies the character, the look, and the background; the video supplies the movement — gestures, body language, head turns, dance moves. An optional text prompt lets you steer the scene, mood, camera movement, or details you want preserved.
It runs in two orientation modes. Follow video angle (the default) matches your character to the camera angle of the reference footage and accepts reference videos up to 30 seconds. Follow image angle keeps the perspective of your uploaded image and accepts reference videos up to 10 seconds. Reference images can be JPG, PNG, or WebP up to 10 MB; reference videos can be MP4, WebM, or QuickTime, and work best when the head and upper body — or the full body — are clearly visible. A Keep original sound switch carries the reference video's audio into the output when it's available.
On Prizmad, Kling Motion Control 3 sits in the same AI Studio as image generation, AI avatars, voiceover, and music — all using one token balance. Generations run asynchronously with an estimated time of about three minutes, and finished clips land in your asset library ready to drop into the video wizard.
Record one talking or gesturing performance, then transfer it onto different AI-generated presenters — new face, outfit, and background per ad variant, same convincing motion.
Take a trending movement pattern from your own reference footage and put your brand character or mascot through the same moves for platform-native creative.
Animate an illustrated or generated brand character with genuinely human motion instead of stock animation loops — one image plus one performance clip.
Keep the winning performance constant and swap the character image across demographics, styles, or outfits to test which presenter converts best.
Capture one clean product-handling motion on video, then reproduce it with different characters or settings drawn from your reference images.
Maps the movement in your reference video — gestures, body motion, head turns — onto the character from your uploaded image, so one performance can drive any character.
Follow video angle matches the reference footage's camera perspective and supports reference clips up to 30 seconds; follow image angle preserves your image's perspective with clips up to 10 seconds.
A single clear character image with visible proportions is enough — the model reads identity, outfit, and setting from it. JPG, PNG, or WebP up to 10 MB.
Add a text prompt to describe the scene, action, mood, camera movement, or specific details to preserve — or leave it empty and let the reference video do the talking.
Toggle Keep original sound (on by default) to bring the reference video's audio track into the generated clip when it's available — useful for trend sounds and spoken performances.
Generations queue and render in roughly three minutes, so you can line up several motion transfers back to back while earlier ones finish.
Open the Kling Motion Control 3 workspace in AI Studio.
Upload your reference image — a clear character shot with visible proportions. This defines who appears in the clip, what they wear, and what the background looks like.
Upload your reference video — a realistic clip where the head and upper body, or the full body, are visible. This defines the motion.
Pick a character orientation: Follow video angle for reference clips up to 30 seconds, or Follow image angle to keep your image's perspective with clips up to 10 seconds.
Optionally write a prompt describing the scene, mood, camera movement, or details to preserve, and decide whether to keep the reference video's original sound.
Click Generate. The clip renders in about three minutes and appears in your asset library, ready for the video wizard or direct download.
Cost depends on the orientation mode: Follow image angle costs 337 tokens per generation (reference video up to 10 seconds), Follow video angle costs 1,008 tokens (reference video up to 30 seconds). The workspace shows the cost before you click Generate.
Kling Motion Control 3 is available from your Prizmad token balance alongside image generation, AI avatars, voiceover, and music — no separate Kling account, API key, or per-provider billing to manage.
If your tokens run out mid-campaign, buy a one-off top-up directly from the top-up page. Generated clips are yours to use commercially.
No — it's a token-based tool on Prizmad. Follow image angle mode costs 337 tokens per generation and Follow video angle mode costs 1,008 tokens. Tokens come from your Prizmad token balance, and you can buy one-off top-ups if you run out. There's no separate Kling account or billing.
337 tokens in Follow image angle mode (reference video up to 10 seconds) or 1,008 tokens in Follow video angle mode (reference video up to 30 seconds). The cost is fixed per generation within each mode, and the workspace displays it before you generate.
Follow image angle keeps the camera perspective of your uploaded character image and accepts reference videos up to 10 seconds. Follow video angle (the default) matches the character to the camera angle of the reference footage and accepts reference videos up to 30 seconds. Use image angle when your image's framing matters most; use video angle for longer performances or when you want the output to feel like the original footage.
Two uploads: a reference image (JPG, PNG, or WebP up to 10 MB) with a clear character and visible proportions, and a reference video (MP4, WebM, or QuickTime) where the head and upper body or the full body are visible. A text prompt is optional.
It can. The Keep original sound switch is on by default and carries the reference video's audio into the generated clip when it's available. Turn it off if you plan to add your own voiceover or music from Prizmad's audio tools instead.
The workspace estimates about three minutes per clip. Generations run asynchronously, so you can queue the next one while the current one renders.
Yes. Clips you generate with Kling Motion Control 3 on Prizmad are yours to use commercially — in paid social campaigns, on landing pages, and on owned channels. Make sure you have the rights to the reference image and video you upload.
Standard image-to-video tools invent motion from a prompt — you describe movement and the model improvises it. Motion Control 3 copies real motion from footage you provide, which gives you precise, repeatable performances. If you just want to animate a static product shot with a camera move, Prizmad's Video from Image tool is the simpler (and cheaper, 3-token) option.
Published 2026-07-16 · Last updated 2026-07-16