TikTok Cross-Border E-Commerce Image Batch Generation

Problem

Sellers listing on TikTok Shop often need to turn ERP product sheets, hero images, and SKU photos into product hero images, lifestyle scenes, and detail assets. Manually cleaning lists, writing prompts, and adapting copy for multiple marketplaces can introduce wrong claims, inconsistent SKUs, or unsupported languages.

How It Works

The skill organizes SKILL.md, Python scripts, and templates into a batch image pipeline:
- Spreadsheet input: supports Miaoshou ERP exports such as .xlsx, .xlsm, and .csv, grouping tasks by product ID.
- Reference image understanding: reads hero images first, then SKU images, with up to 10 reference images per product.
- Template matching: selects from built-in e-commerce templates for hero shots, lifestyle scenes, detail macros, infographics, and packaging.
- Language adaptation: uses the target site to choose Malay, Vietnamese, English, Japanese, Portuguese, and other supported languages; defaults to Chinese when unspecified.
- Batch execution: submits jobs in batches, polls status, downloads results, and writes per-product logs and a manifest for recovery.

Boundaries

It is best for batch generation from clean product photos and structured lists, not for delivering concept renders as real product assets. The per-batch limit is 2000 images and must be split into separate submissions; output volume also depends on account quota, service limits, and task complexity. Brand names, certifications, packaging text, and on-image copy should be verified manually before use.