AI Motion Control API

An AI motion control API transfers movement from a reference video to a character image. It generates a new clip combining visual identity and performance information. Model-specific controls determine how the character, camera and scene are handled.

The motion control models in this task use a character image and a source video. Read the endpoint for exact fields, accepted formats, size and duration limits. Optional prompts or orientation controls should only be supplied when that model documents them.

Use one clearly visible primary subject performing the intended action. Prefer stable framing and limited occlusion, and avoid unrelated cuts. Select movement that suits the character image and test a short segment before sending a longer performance.

A compatible starting pose, facing direction and crop can reduce ambiguity. A close-up portrait is a poor reference for a full-body action with complex leg movement. Align the amount of body visible and use the orientation settings documented for your model.

Support and quality depend on the selected model and source assets. Start with a clearly drawn subject and a compatible performance. Test stylized anatomy, hands and fast movement carefully rather than assuming the model handles every character type equally.

Large pose differences, hidden limbs, fast motion and unclear source details can make transfer harder. Use a clearer character image, simpler performance and compatible framing. Change one input at a time and compare a short output to identify the problem.

Do not assume multi-person transfer. The selected endpoint may focus on one primary subject, and overlapping people can make motion assignment ambiguous. Confirm documented subject limits and begin with a single-character reference when testing.

Movement transfer and speech synchronization are different requirements. Check the model audio behavior and do not assume that a moving mouth follows the soundtrack. Use Talking Avatar or a documented lip-sync workflow when accurate speech is the main goal.

Submit the model and documented character and video inputs through PoYo generation. Retain task_id, then poll status or use a supported callback_url. Inspect the completed clip for motion fidelity, identity and cropping before delivering it to users.

Read the selected tier to determine which duration is billed and whether input and output processing are counted. Quality modes can have different rates. Estimate the cost from the reference length and documented billing rules rather than treating every model as a fixed-price clip.

Motion Control uses a video performance to guide a character image. Image to Video usually starts with an image and prompt-led movement. Choose motion transfer for a specific gesture or dance, and ordinary image animation for camera moves or simpler actions.

Test representative crops, facing directions, clothing, expressions and movement speeds. Verify file limits and price units, then track task status and failures in your application. The model page and llms.txt resource provide the current request structure for integration.