What Motion mode does, and when to use it

Three of Creative Studio's four modes work roughly the same way: describe something, attach references, generate. Motion doesn't, and the differences aren't cosmetic.

What does Motion actually do?

Transfers kinematic movement from one clip onto a different character.

You supply a driving video — footage of whatever movement you want — and a character image. The output is your character performing that movement.

So the creative decision isn't "what should happen," it's "what should this look like while it happens." The motion is already determined by the time you start.

That's a different mental model from Create, where you describe an action and the model invents the movement. Here you've brought the movement with you.

In Create you describe the action. In Motion you supply it. The prompt does something else entirely.

What does it need?

Two inputs, and neither is optional.

The generate button stays locked until both inputs validate. That's unusual — the other three modes let you generate with a prompt alone.

The character image requirement is more specific than it looks. Face and upper body, clearly visible. A full-body shot at distance or a tight close-up on the face both give the model less to map movement onto, since the transfer needs to understand where the shoulders and torso are.

And the driving video comes from your workspace library, not from an arbitrary upload — so it's a clip you've already generated or brought in.

Why must the prompt exclude motion?

Because the motion is already supplied, and describing it again creates a conflict.

Hexcoded's own documentation is direct about this: all action dynamics come from the driving video, and conflicting motion verbs in the prompt can cause visual jitter.

So a prompt reading "she walks forward and turns" competes with a driving video that already contains walking and turning. The model has two sources for the same information and they don't agree.

What the prompt is for instead: environment, background, lighting and tone. Where the movement happens and what it looks like — not what the movement is.

Every motion verb you write is a second opinion the model didn't ask for. Describe the room, not the action.

A dim apartment kitchen at night, single overhead bulb, cool blue light from a window behind. Worn wooden surfaces, cluttered counter. Muted colour, slight grain.

Where this falls short. "No motion verbs" is easy to state and harder to write. Descriptions naturally reach for action, and a prompt that says "leaning against the counter" has smuggled a pose in. The test is whether a sentence describes the space or the person.

How is this different from the other three modes?

Four ways, and each one removes a decision you'd make elsewhere.

One pipeline, not a picker. Create offers eighteen models, Edit nine, Extend seven. Motion runs on a single dedicated pipeline — Kling 3.0 Motion Control — so there's no model choice to make.

No duration control. Length is inherited from the driving clip's exact runtime. No slider, no range.

Aspect ratio inherited too. Like Edit and Extend, Motion takes framing from the source and hides the selector.

Two mandatory inputs. The other three modes will generate from a prompt and whatever you've attached. Motion won't start without both.

What is it actually for?

Three cases where supplying the movement beats describing it.

A specific physical action you can't prompt reliably. A particular gait, a gesture with exact timing, a piece of business that has to look a certain way. If you can film it or find it, transferring it is more reliable than describing it.

The same movement across different characters. One driving video, several character images, and you get the same action performed by each. For a series with parallel scenes or a campaign with several presenters, that's a consistency route nothing else offers.

Reusing a take whose performance worked but whose casting didn't. The movement survives; the face changes.

That second one is the underused case. Most people reach for Motion when a prompt failed, rather than when they want repeatability.

When shouldn't you use it?

Three cases, and the first is the most common mistake.

When you don't already have the movement. Motion needs a driving video. If you'd have to generate one first, you've added a step rather than removed one — generate the shot directly in Create instead.

When the character needs to speak. Motion transfers kinematic movement. A face delivering a script is Talking Actors, where duration comes from the script and emotion control places a beat at an exact word.

When the environment is doing the work. Motion's prompt covers environment, but the mode exists to move a character. If the shot is about the space rather than the person, Create gives you more control over it.

How should you actually use it?

Five steps, and the order matters because two of the inputs constrain everything downstream.

Motion mode behaviour described here comes from Hexcoded's Video: Motion documentation and the Creative Studio picker as observed in September 2026. Pipeline, input requirements and behaviour change with versions. Verify in the picker before relying on this for a commercial delivery.

More on model capability, access and rights in Models.