How Kling Motion Control is billed, and how to budget it

Updated 2026-10-02

Motion Control is the Kling model family that transfers the movement from a reference video onto a character in a still image. Its billing is different enough from ordinary text-to-video that budgeting it like a normal Kling clip leads to wrong numbers: the output length is not something you simply type into a duration_secs field. This page explains what VideoRouter's documentation says about how it is priced, how the two input forms change the cost, and how to budget a project. Rates are shown live, not typed here, because they change.

What is documented

Motion Control runs through the same POST /videos endpoint as every other model; there is no separate endpoint. VideoRouter's documentation lists two tiers with separate model ids, kling/motion-control-std/novita and kling/motion-control-pro/novita, each with its own per-second rate, and shows a single host serving them, with no cross-provider failover pool. The docs also state that both input orientations bill the same per-second rate regardless of which field carries the reference video, and that the displayed cost already includes the 2% platform fee. The live rates sit on the cost page and the pricing page.

Per second, before VideoRouter's 2% platform fee. For tiered models each row compares the resolution tier with the widest host-to-host gap. Built 2026-10-02 from the live catalog.

Find the Motion Control rows there for both tiers. Hosts can change, so check the model page for the current list before relying on a single-host assumption.

The two input forms and what each does to length

The only thing that changes the billable seconds is how you supply the reference video:

So with the first form your cost per output is roughly the tier rate times a fixed short length, and with the second form it scales with the length of the reference clip you upload. Same rate, different seconds. That is the whole pricing model, and it is why the input form is the main budget decision after the tier.

Budgeting method

Use symbols, then fill them from your own logs and the live table. Let m be the tier's per-second rate and L the billable seconds per job.

cost_per_job   = m * L
L (references form)  = the fixed short output length from the docs
L (input_video_url)  = probed length of the reference clip, capped at 30 s
cost_per_accepted    = cost_per_job * attempts_per_accepted

Three practical consequences follow:

1. Trim the reference before you upload. With input_video_url, a reference with idle lead-in and tail seconds produces output you pay for and then cut. Cut the reference to the motion you want and your L drops with it.
2. Prototype in the short form. The fixed-length orientation caps the cost of a test render. Check that the character image and reference framing work before committing to a long render.
3. The 30-second ceiling is your worst case. A single job's cost is bounded by the tier rate times thirty seconds. Use that as the per-job cap when you set a key's monthly spend limit.

Worked layout for a project, symbolic

Say you need N finished motion clips, each of reference length L_i. Budget in two phases. Phase one is exploration: E short test renders in the fixed-length form on the Std tier, costing E * m_std * L_short. Phase two is production: for each of the N clips, a attempts on the tier you chose, using input_video_url, costing sum over clips of (a * m * L_i). Add the phases, and keep the 2% fee out of the formula if the rates you read already include it, as the docs say they do. The calculator handles the arithmetic for ordinary clips; for Motion Control, the hand formula above is what matters because the length is driven by the reference.

Std versus Pro for Motion Control

The tiers follow the same pattern as Kling v3.0 generally: separate ids, separate rates. The reasoning in the tier price-gap guide carries over: Pro is worth its ratio only if it reduces your attempts per accepted clip by more than that ratio. For Motion Control the reference clip often matters more than the tier for how the result looks, so fix the inputs on Std first and only then test Pro.

Things the documentation does not tell you

• This page does not state a minimum reference length or exact output seconds for the fixed form beyond "about four seconds"; test one job and read the charge.
• It does not claim any quality difference between tiers.
• It does not describe a refund path for outputs you dislike. A completed job is billed; only jobs that fail because every upstream host failed are not.

The most reliable verification is cheap: run one short job per form with a capped key and compare the dashboard charge with m * L. Mechanics and field details are in the Motion Control docs; create a key to try it, and see the cost-per-minute method for turning job costs into project budgets.

Frequently asked questions

Keep reading

• How to Reduce Your Kling API Cost — Five Practical Levers
• Kling API Cost per Minute: Budget Formula for Real Projects
• Kling vs Seedance vs Veo Pricing: Compare Per-Second Fairly
• Why Kling API Prices Differ by Provider (and How to Choose)