Kling motion control prompt: what it can change

In Kling motion control the prompt does not describe the movement: the reference video supplies that. Kling says you enter a prompt to control background elements and other information, and its examples describe clothing and an extra animal in the scene.

This is from Kling's own Motion Control User Guide, read 2026-09-29. The Sume field is from the Sume API reference; see also the API reference docs.

What is the prompt for in Kling motion control?

The guide's steps for 2.6 end with entering a prompt to control background elements and other information. It also says movements, expressions and camera movements follow the motion reference, with other details controllable through prompts.

The showcase prompts are short, plain descriptions of the scene, not shot lists:

• "A girl wearing a white tank top and a denim skirt."
• "A corgi runs in, circling around a girl's feet."

What does the prompt do on Sume?

prompt is optional on POST /v1/kling/3.0/motion-control, from 1 to 2,000 characters. Sume's field description says motion and framing follow the driving video and that the prompt only steers appearance details.

So on Sume, describe how the character and scene should look, not how they should move. Kling's corgi example adds a new subject; Sume's description talks only about appearance details, so treat added subjects as something to test on a short clip.

curl -X POST https://api.sume.com/v1/kling/3.0/motion-control \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: motion-prompt-001" \
  -d '{
    "image_url": "https://example.com/girl.png",
    "motion_video_url": "https://example.com/spin-5s.mp4",
    "duration_seconds": 5,
    "prompt": "A girl wearing a white tank top and a denim skirt."
  }'

What should I put in the prompt, and what should I leave out?

These are suggestions built from the two descriptions above, not rules from either page.

• Put in: clothing, appearance and the background, the kinds of detail Kling's own examples describe.
• Put in only what the still does not already show; the image is already the starting appearance.
• Leave out: instructions for how the character should move. The reference video supplies the movement, so a movement line has nothing to steer.
• Keep it short. Kling's examples are one sentence each, and Sume's field stops at 2,000 characters.

Can I change the movement with the prompt?

No. Kling describes motion control as assigning motion, extracted from an uploaded video, to a character in your image, and Sume states that the output follows the driving video's motion. To get different movement, use a different reference video. Motion control API: animate an image with a driving video lists the other fields.

What are the prompt limits?

Sume's schema sets the limits; Kling's page states none.