AI Multi-Region Product On-Model Studio

The Core Pain Point in E-commerce Localization

For cross-border and multi-regional retail, the same garment requires visual adaptation to suit local aesthetics, body types, and scenarios. Traditional methods rely on multiple真人 photoshoots, which are costly and slow. The core problem is: How to rapidly generate a market-specific, model-on-product e-commerce image from a single confirmed product master shot, while preserving 100% of SKU details (silhouette, color, material, hardware, etc.)?

How the Skill Works: From Product Confirmation to Market Delivery

The skill automates the workflow via these steps:

1. Confirming Inputs and Directions

   • Requires three inputs: a confirmed product master image (images[0]), the target market, and model direction (pose, outfit, scenario).
   • Product details, model identity, and market facts are not assumed known; they must be provided or clarified by the user.

2. Core Image Generation and Editing

   • New Image Generation: Uses beatra.images.transform to create a new model-on-product image based on the product shot.
   • Localized Edits: For tweaks on an existing result, uses beatra.images.edit, placing the original result into images[0] as the trusted base image.
   • Before invocation, queries available models via beatra.models.list (e.g., image_to_image or image_edit capabilities), using defaults like model: "auto" and count: 1.

3. Preview and Paid Execution

   • Adheres to a strict "approve first, then execute" principle. Before calling, it presents the complete product info, market direction, generation prompts, canvas parameters, model selection, and real-time pricing to the user.
   • Paid tasks are initiated via the bundled client (scripts/mcp_client.py with the call command) only after explicit user approval.

4. Task Tracking and Result Delivery

   • Immediately records the task_id after creation and polls it to terminal state (success/failure) using beatra.tasks.get.
   • Upon delivery, explicitly reports the output file, dimensions, model used, and actual credits charged (billing.net_charged_credits), and verifies SKU fidelity and market adaptation.

Key Boundaries and Considerations

• Does Not Process Unconfirmed Inputs: This skill is not for new product shoots or generating pure product images (use product-photo-studio). All product, market, and model information must be confirmed before execution.
• Strict Idempotency and Recovery: Every new request (any change in input) requires a new client_request_id and user approval. If a task ID is lost, recovery is possible via beatra.tasks.list and beatra.tasks.get, but auto-replaying unapproved requests is prohibited.
• Paid Operations and Cancellation: Task cancellation must use beatra.tasks.cancel and track the original task status. No alternative paid requests are created automatically.
• Technical Limitations: All communication is solely via the bundled mcp_client.py. Directly invoking the host Beatra Connector or using REST/OpenAPI as a fallback is forbidden.