Explainer Video through MCP
Explainer Video uses explainer_video and creates continuous animated scenes with narration.
Call kp_get_catalog first, then pass these arguments to kp_create_creation:
{ "type": "explainer_video", "prompt": "Explain MCP with continuous animated scenes", "language": "en", "aspect_ratio": "16:9", "duration_seconds": 30}
duration_seconds belongs only to Explainer Video and must follow the current catalog limits. Select voice_id, style_id, and output_quality from the same catalog when needed.
After creation, poll kp_get_job_status, then read the result with kp_get_creation. For a supported scene, use kp_edit_creation with update_scene_narration or regenerate_scene.
When to choose Explainer Video
Choose this format when the viewer needs one continuous explanation: a product walkthrough, a training concept, a technical mechanism, or a decision that benefits from narration and connected scenes. Put the audience, the source facts that must remain accurate, and the intended viewer takeaway in the prompt. A prompt such as “explain our login policy” is less useful than one that names the audience, the sequence, and the terms that must be preserved.
Run the job to completion
Creation is asynchronous. Keep the job_id returned by kp_create_creation, poll it with kp_get_job_status, and only present the result after the status succeeds. Then use kp_get_creation as the source of truth for main_url, artifacts, scenes, and editable_actions. Do not derive download paths or assume that a queued job has a playable result.
If the job fails, read the structured status before retrying. Correct unsupported catalog values or an incomplete prompt first; retries should represent a changed request, not a blind duplicate.
Continue from here
Quickstartfor the full MCP setup and lifecycleAvailable Toolsfor the current tool surfaceRead Aloud Videowhen the explanation should be page-by-page instead
