Does the Kling motion control prompt change the movement? No
On Sume's Kling 3.0 Motion Control route the motion video drives all movement and the prompt only steers appearance. Fields, limits and $0.1575 per second.

No. On Sume's Kling 3.0 Motion Control route the motion reference video drives all of the movement, and the optional prompt only steers appearance. If you want the character to turn, jump or gesture differently, change the reference video, not the prompt (as of 2026-10-09).
The request, field by field
The route takes exactly one visual source, either an image_url or an avatar id or handle, plus a motion_video_url and a duration_seconds. Everything else is optional.
| Field | Required | Role |
|---|---|---|
| image_url or avatar_id or avatar_handle | One of them | The character still |
| motion_video_url | Yes | Drives all output motion |
| duration_seconds | Yes, 1 to 30 | Reservation basis; the reference length sets the output |
| prompt | No | Appearance steering only |
| keep_original_sound | No, default true | false returns a silent clip |
| character_orientation | No, video or image | Default video |
What this means for prompts
Write the prompt about how the character looks in the result: clothing, lighting and surroundings. Leave out verbs that describe the movement, because the reference already supplies it. The route also rejects Video 1.0 vocabulary such as reference_image_urls, so you cannot widen it with fields that belong to the text-to-video model.
The price follows the declared duration, not the prompt: ceil(duration_seconds) x $0.1575. A 7.2-second reference is 8 seconds, or $1.26. A 15-second reference is 15 x 0.1575 = $2.3625.
Common mistakes
The first mistake is a prompt like walk toward the camera when the reference video shows someone standing still. The result follows the reference. The second is sending a long reference when only five seconds matter: you pay for the whole declared duration at $0.1575 per second, so a 20-second reference is $3.15 and a trimmed 5-second one is $0.7875.
The third is mixing vocabulary from other routes. The Sume route rejects Video 1.0 fields such as reference_image_urls, and it rejects a model field in the body, because the URL is the model. If a request returns invalid_request, remove everything the table above does not list.
Checklist before you submit
Trim the reference to the movement you want, since its length is the clip length. Use a still where the character faces the camera in a pose that resembles the reference's first frame. Decide on sound: keep_original_sound defaults to true, so the reference's audio carries over unless you set it to false. Through MCP the tool is kling-motion-control_create, a paid tool that needs the Write grant, listed on MCP tools and gates.
For a still with a ready avatar, pass avatar_handle in place of image_url. Only one of the visual sources is allowed in a single request.
Finally, character_orientation takes video or image and defaults to video. Pick it deliberately and note it with the job, so that a second run with the same inputs differs only in the field you meant to change.
Sources
Related posts
More in Media tools
- Eight inspect stills on a 45 s Snap ad: one lands in the first 6 s
Video inspect's default eight stills on 45 seconds sit 5.6 s apart; only 2.8 s falls in Snap's non-skippable first 6. Use frames at 1, 3 and 5.9 instead.
- VEED Fabric on a 41.5-second voice note: $4.20 at 480p, $7.88 at 720p
A 41.5-second voice note bills 42 seconds on Sume's VEED Fabric 1.0: 42 x $0.10 = $4.20 at 480p, 42 x $0.1875 = $7.875 at 720p, reserved as $7.88.
- Fabric accepts image_url or avatar_handle, never both: which to send
On Sume's VEED Fabric 1.0 route, send one visual source: the image_url of an inspected still, or an avatar_handle the user named. Both together is not allowed.
- Fabric vendor $0.15 per second is $0.1875 on Sume: five lengths
fal lists VEED Fabric 1.0 at $0.15 per second (720p) and $0.08 (480p). Sume bills list x 1.25: $0.1875 and $0.10. Prices at 4, 8, 12, 20 and 30 seconds.
Written by Sume