Test a Kling motion control look on 10 seconds for $1.575, then run 30

Before you pay for a full 30-second Kling 3.0 Motion Control run, cut the first 10 seconds of the reference and run that. On Sume it costs $1.575 against $4.725 for the full clip, so a test that shows a bad still, a wrong character_orientation, or a prompt that fights the motion saves you $3.15 per miss.

A test cannot prove the whole 30 seconds will hold. It tells you about identity, framing, and the first stretch of motion, which is where most setup mistakes show up.

What each run costs

The rate is ceil(duration_seconds) times $0.126 times 1.25, which is $0.1575 per second. Both runs are reserved at admit and captured on completion. A failed job is refunded.

Cut the test with video trim

POST /v1/video-trim copies a [start, end) range of a clip already on Sume into a new MP4 for $0.02 per job, and returns a media.sume.com URL you can use as motion_video_url. Use the start of the clip for the test, or the part with the fastest motion if you fear that part most. Pass duration_seconds: 10 on the test job and the full measured length on the final.

What to look at in the test

If the test drifts in the first second, change the still before you spend the final. Kling's action videos run 3 to 30 seconds (read 2026-10-07), so a 10-second cut is a valid reference on its own.

• Identity: does the face and clothing in the output match the still?
• Framing: character_orientation set to video keeps the driving video's framing, image keeps the still's. Pick one on purpose.
• Audio: keep_original_sound defaults to true, so the reference audio is in the output unless you turn it off.
• Prompt: motion and framing follow the video, the prompt only steers appearance. If you see no effect, that is by design.

When a test is not worth it

For a reference of 6 seconds the test is most of the final, so just run it. A test pays when the reference is long and the result matters. If you plan more than one still against the same motion, test each still on the 10-second cut, then run the winner on the full clip. For the rate itself, see the per-second price and a 30-second clip.