modelstringFixed model ID. Swap it with kling-v3-motion-control to switch versions.
Default: kling-v2.6-motion-control
Transfer a reference video's motion to one character image at 720p or 1080p. Use EvoLink's unified API to evaluate dance clips, avatar gestures and motion previews, then choose whether 2.6 meets your brief or 3.0 is worth evaluating.
Upload a character image and a motion video, pick the orientation and resolution, and preview the result before you wire up the async API.
Create a Sample
Add a character image and a motion video, then set the parameters.Required. One JPG or PNG up to 10 MB showing the character's upper body or full body.
Required. One MP4 or MOV up to 100 MB. Use 3–10 s with Follow image or 3–30 s with Follow video. Complex or fast motion may produce a shorter video.
The character keeps the facing of your image. The motion video can be up to 10 s.
On: the output uses the motion video's own audio. Off: the output is silent. No new audio is generated.
Your motion-controlled video will appear here
Add a character image and a motion video, then click Generate.
Results are kept for 24 hours — download anything you want to keep.
Pay per second of actual output video. 1080p uses 1.6x the 720p rate; complex or fast motion may produce a shorter output than the reference.
First Test Cost
3s motion video - 720pOne 3s 720p clip, the shortest motion video allowed
About 50 tests with $10 in credits.
In the playground, upload a 3-second motion video and keep 720p to reproduce this minimum-cost test. Trim longer clips to the section you want to test.
Testing Budget Guide
Pick an amount based on how many tests you expect.Good for a first validation.
Good for trying different characters and motions.
Good before production integration.
Estimated total for one run, with the reference length rounded up. Final billing uses the actual output length, rounded to the nearest whole second.
kling-v2.6-motion-control| Item | 720p | 1080p | Billing |
|---|---|---|---|
| Motion controlOutput video, per second | $0.067/s-10% 4.5 cr/s$0.074 Official price | $0.106/s-10% 7.2 cr/s$0.118 Official price | Output seconds (follows the motion video) |
Estimates round reference seconds up; final billing rounds actual output to the nearest second. 1080p = 720p x 1.6. Submission reserves credits for the reference length (10 s if unreadable); settlement refunds any unused reserve.
Kling 2.6 Motion Control transfers a reference video's performance onto one character image. Use the same EvoLink video endpoint as Kling 3.0 Motion Control: change the model ID and remove element_list when switching to 2.6. Its rate is about 55% of 3.0 at 720p and 66% at 1080p. Output generally follows the reference length, but complex or fast motion may shorten it. Original audio can be kept or muted; this route does not generate new audio.
kling-v2.6-motion-controlOne model ID for image-plus-video motion transfer. The request shape is identical to kling-v3-motion-control; only subject elements (element_list) are not available.
These fields cover the main integration decisions. Open the API tab for the full request and response schema.
modelstringFixed model ID. Swap it with kling-v3-motion-control to switch versions.
Default: kling-v2.6-motion-control
image_urlsarrayOne character image: JPG or PNG, up to 10 MB, 300-65536 px per side, aspect ratio 1:2.5 to 2.5:1.
Default: required
video_urlsarrayOne MP4 or MOV up to 100 MB; 3–10 s with image orientation or 3–30 s with video orientation. Output may be shorter.
Default: required
model_params.character_orientationenumimage keeps the image's facing (video up to 10 s); video follows the clip (up to 30 s).
Default: required
qualityenum720p (std) or 1080p (pro, billed at 1.6x). 4k is rejected.
Default: 720p
model_params.keep_soundbooleanKeep the soundtrack of the motion video; false mutes the output.
Default: true
promptstringAdd details such as clothing, scene elements, or camera effects.
Default: optional
callback_urlstringHTTPS webhook that receives the task result instead of polling.
Default: optional
Record or pick a clip of the movement you want, then apply it to any character image instead of describing complex choreography in a prompt.
The soundtrack of the motion video is carried into the output, so dance and performance clips stay in sync with their music. Turn it off to get a silent video.
The request and response match Kling 3.0 Motion Control. Start on 2.6 for volume and move individual jobs to 3.0 when you need subject elements.
Use the EvoLink API for product integration and batch jobs, or call it from Codex, Claude, or Gemini for fast creative and development workflows. Both paths share the same EvoLink API key, balance, model routes, and task history.
Best for: product backends, batch jobs, automated workflows
Call EvoLink’s unified video API from your server and control the model ID, parameters, task queue, callbacks, and result storage.
Best for: creative and development tasks in Codex, Claude, and Gemini
Give the Agent your output goal, assets, and acceptance criteria. It can choose the route, assemble the request, track the task, and return the result without requiring you to hand-code every step.
Start with a full-body character image and a short, continuous dance reference. Check foot placement, limb crossings and timing before reusing the workflow for other characters. A completed task still needs a visual review.
Pair an upper-body gesture with an image showing the same body area. Review face, clothing and silhouette through the whole clip. Test head turns and hands crossing the face separately before choosing a version.
Use a short reference to check a character's movement and framing before committing to a longer shot. For actions involving a chair, bicycle or other prop, inspect contact and object shape; treat these as tests, not guaranteed production results.
This example shows the shortest runnable flow: create a task, save the returned task ID, then poll or wait for an HTTPS callback. Open the API tab for the complete parameter and response reference.
curl -X POST https://api.evolink.ai/v1/videos/generations \
-H "Authorization: Bearer $EVOLINK_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "kling-v2.6-motion-control",
"image_urls": ["https://example.com/character.png"],
"video_urls": ["https://example.com/dance.mp4"],
"quality": "720p",
"model_params": {
"character_orientation": "image",
"keep_sound": true
},
"callback_url": "https://your-domain.com/webhooks/video-task-completed"
}'
# Save the returned task ID, then query:
curl https://api.evolink.ai/v1/tasks/{task_id} \
-H "Authorization: Bearer $EVOLINK_API_KEY"Check every value against the parameter cards above: ranges, allowed options, asset counts, and file types must match the selected route.
Check the Authorization bearer token and confirm the available balance and task billing in the console.
Review likeness rights, brand assets, readable text, sensitive content, and input-asset compliance.
A Kling 2.6 Motion Control task that reaches a final failed state is not charged. Keep the task ID, inspect the final state, and identify whether the cause is validation, review, or execution before retrying.
callback_url must be a public HTTPS URL, not localhost or a private network address. Always keep the task ID as a polling fallback.
Choose against your own acceptance criteria: motion timing, character identity, framing and usable duration. The guidance below is an evaluation plan, not a measured quality ranking.
| Feature | Kling 2.6 Motion Control | Kling 3.0 Motion Control |
|---|---|---|
| Model ID | kling-v2.6-motion-control | kling-v3-motion-control |
| Price per output second | About 55% of 3.0 at 720p, 66% at 1080p | Higher |
| 1080p multiplier | 1.6x the 720p rate | 1.334x the 720p rate |
| Subject elements (element_list) | Not supported | Up to 1 element |
| Inputs, orientation and duration rules | Same as 3.0 | Image + video, 3-30 s, Follow image or video |
| When to choose it | Cost-sensitive volume and general motion transfer | Jobs requiring subject elements |
Need subject elements? See Kling 3.0 Motion Control →
EvoLink sends your request to Kling's motion control endpoint with the 2.6 model. Parameters are validated before the task is queued, and element_list is rejected with a 400 because 2.6 does not support subject elements.
Submit a task, then poll GET /v1/tasks/{task_id} or receive a callback. The finished video URL is kept for 24 hours, so copy it to your own storage.
Credits are reserved for the reference length (10 s if unreadable). Estimates round up; final billing rounds actual output seconds to the nearest whole second, at 1.6x for 1080p. Any unused reserve is refunded.
See the per-second rate for 720p and 1080p and estimate each run before you generate.
Call Kling and other video models through one API key and one request format.
Manage credits, spending, and model usage in the same EvoLink account.
Track pending, processing, completed, and failed tasks through a consistent async workflow.
Change the model ID to choose 2.6 or 3.0. Remove element_list when switching to 2.6; use 3.0 for subject elements.
Compare Kling routes under one EvoLink account and API key, then choose by capability, quality, and cost.


Kling 3.0 Motion Control
Motion transfer with subject elements for more consistent characters.
View API

Seedance 2.0
A video model for text, image, and reference-driven generation with native audio.
View API


It is Kling's motion control model at version 2.6. You give it a character image and a motion video, and it returns a video in which the character performs the motion from the clip.
Use kling-v2.6-motion-control. Kling's official API names the same model kling-v2-6.
The endpoint and orientation limits are shared. 2.6 is about 55% of the 3.0 rate at 720p and 66% at 1080p, but rejects element_list. Use 3.0 for subject elements and evaluate both on your own references.
Yes. Change the model field between kling-v2.6-motion-control and kling-v3-motion-control. Remove element_list when you send a request to 2.6.
No. Requests that include a non-empty element_list are rejected with a 400 error. Use Kling 3.0 Motion Control when you need subject elements.
At least 3 seconds; image orientation allows up to 10 seconds and video up to 30 seconds. Output generally follows the reference, but complex or fast actions may reduce the usable duration.
image keeps the facing and framing of your character image, which suits camera-facing performances. video makes the character turn the way the person in the clip does, which suits full-body moves and turns.
Match full-body or upper-body framing between your image and video. Keep the head and limbs visible and use one continuous shot with moderate movement. Avoid cuts, occlusion and fast camera moves; test a short clip before running a batch.
It is billed per second of output video, and 1080p costs 1.6x the 720p rate. Use the live pricing section and the playground estimate for the current rate of your account.
Estimates round reference length up. At submission, credits are reserved for the reference length (10 s if unreadable). The final charge rounds actual output length to the nearest whole second; any unused reserve is refunded.
No. The prompt is optional and can add details such as clothing, scene elements, or camera effects.
Yes, by default. Set model_params.keep_sound to false to mute the output.
EvoLink result URLs are available for 24 hours. Copy completed videos to your own storage instead of using the temporary URL in production.
This route transfers motion from a reference video to one character image. It is a separate workflow from ordinary text-to-video or image-to-video generation and does not provide a brush editor for drawing motion paths. Choose the route by its inputs, not just the Kling version number.
Do not assume the source background will be preserved frame by frame. Motion comes from the video, while the character image provides the appearance and scene reference; a prompt can influence scene details. If the original background must stay unchanged, evaluate a video-editing workflow against that requirement.