Skip to main content
Replace characters in an image or video with reference characters.For Akool video replacement, analyze the video first, then pass the completed analysis ID and person mappings. For Kling video replacement, provide image and the video in target_url; the scene must contain a single person. Character replacement preserves the scene while changing its character.

​
API Endpoints

​
Getting Started

  1. List the models available for this feature.
  2. If required by your model, analyze the source and wait for the analysis to complete.
  3. Create a task using the model-specific parameters.
  4. Poll task results or receive the final webhook notification.

​
Billing

Generation is priced using the selected model configuration and your account/team pricing. Image generation is charged per task; video pricing also depends on duration and resolution. Credits are reserved before queueing and finalized through the existing generation billing flow. Creation errors do not produce a billable task. Read deduction_credit per created task rather than treating the sample value as a price quotation. AIGC website membership discounts do not automatically apply to OpenAPI; account and team pricing can change the final credit amount.

​
Response Code Description

A business code other than 1000 indicates that the API request failed. For batch creation, also check data.errors.
See the Webhook guide for callback verification and the complete error list for other account errors.

​
Callback

If webhookUrl is supplied, final task notifications follow the existing encrypted and signed Webhook format. After verification and decryption, read _id, status, type, and url. Fetch task detail to obtain operation-specific fields, including additional hairstyle views. Return HTTP 200 to acknowledge delivery.

​
Input limits

Akool video replacement supports 3–20 seconds and 1–5 distinct analyzed people. Kling character replacement supports 3–30 seconds; image orientation supports at most 10 seconds. Image replacement accepts 1–2 face mappings and selects output size automatically from the target image. For Kling replacement, the selected character model and its configured motion model must both be available to your account. Resolution must be compatible with the configured motion model.