12-second Kling clip: motion control $1.89 vs kling-3 $1.68 or $2.52
For 12 seconds, Sume's Kling motion control costs $1.89, kling-3 costs $1.68 silent or $2.52 with audio. The cheaper one is the one that fits the brief.
A 12-second Kling clip on Sume costs $1.89 on motion control (12 x $0.1575), $1.68 on kling-3 silent (12 x $0.14), or $2.52 on kling-3 with audio (12 x $0.21). The $0.21 gap between motion control and silent kling-3 is small, so decide on what each route can do, not on price.
They answer different briefs
kling-3 writes a scene. You give it a prompt and optionally a first and last frame, and it invents the motion. Motion control copies a scene. You give it a still and a reference performance, and the still moves the way the performer did.
If your brief says a woman walks into a kitchen and opens a fridge, that is kling-3. If it says this mascot should do exactly what this person did in this clip, that is motion control. Choosing by price is a mistake because neither route can do the other's job.
The numbers
Prices are Kling's per-second list on fal times Sume's 1.25. The motion control list is $0.126, and the kling-3 list is $0.112 silent and $0.168 with audio. Each is a per-second card with no resolution tier on the Sume side for these two.
| Route | Per second | 12 seconds | Takes a reference clip |
|---|---|---|---|
| Kling 3.0 motion control | 0.1575 | 1.89 | yes, as motion_video_url |
| kling-3 silent | 0.14 | 1.68 | no |
| kling-3 with audio | 0.21 | 2.52 | no |
Where the length is decided
On kling-3 you pick duration from 4 to 15 and the output has that length. On motion control the reference sets the length, and you declare it in duration_seconds, from 1 to 30. A 12-second reference with duration_seconds: 12 is the correct pairing.
The reserve on motion control is rounded up to a whole second, so a 11.2-second reference costs the same as a 12-second one, 12 x $0.1575. If your reference is a hair over a whole number, trim it to the whole number and save a full second.
A quick decision rule
Ask one question: do you have a performance on video that must be repeated? If yes, motion control, at $1.89 for 12 seconds. If no, and the clip is a scene from words, kling-3. Add audio only if the model's own sound is what you want, since a separate narration or music track is often cheaper and easier to fix than a baked-in one.
Both submit to the same Sume job lifecycle, both reserve at admit and refund on failure, and both accept an Idempotency-Key, so switching between them later is a change of path and body, not a change of pipeline.
Worked example
Say a toy brand wants a 12-second clip of a plush fox waving. With kling-3 the prompt describes the wave, and silent output is $1.68. With motion control the brand films a person waving for 12 seconds, uses the plush photo as the still, and pays $1.89. The second costs $0.21 more and gives the exact wave that was filmed, with the timing of a real person. For a brand that wants repeatable results across a line of plush toys, that is worth more than the 21 cents.
Related posts
More in Comparisons
- 20-second 1080p clip: LTX-2.3 Fast $0.80 vs Wan 3.0 $5.00 on Sume
LTX-2.3 Fast lists $0.04 a second at 1080p, so 20 s is $0.80. Wan 3.0 on Sume bills $0.25 a second at 1080p, so the same 20 s is $5.00. Why the gap?
- 30-second AI video ad: Seedance 2.5 or Wan 3.0, limits side by side
Seedance 2.5 and Wan 3.0 are the two Sume rows that reach 30 seconds. Wan 30 s is $1.88 at 480p, $3.75 at 720p and $7.50 at 1080p; Seedance is billed in tokens.
- 4-second hook at 720p: Wan $0.50, Omni $0.50, Seedance 2.5 $2.31
For a 4-second 720p hook, Wan 3.0 and Gemini Omni Flash 1.1 cost $0.50 on Sume and Seedance 2.5 about $2.31. MiniMax H3 starts at 5 seconds.
- 60-second holiday music video at 1080p: Omni Flash, H3 Max or Wan 3.0
Six 10-second clips at 1080p cost $11.28 on Omni Flash 1.1, $12.00 on H3 Max and $15.00 on Wan 3.0. Add $0.23 for the track and render. Arithmetic inside.
Written by Sume