Amazon A+ Module Stills

Problem it addresses

Sellers working on Amazon A+ Content often mix up main image sets, lifestyle shots, and scrollable detail-page narratives, or fill modules with invented benefits, certifications, and comparison numbers. A+ pages need module stills that read as one coherent page: one visual role per confirmed selling point, with consistent product appearance, color palette, and lighting, not generic template images.

How it works

Core capability is generating Amazon A+ Content module stills from seller-confirmed selling points. The hard inputs are the exact SKU name and benefits already approved by the seller; when key facts are missing, it asks only for SKU specs, seller wording, placement/language, and optional product or brand reference stills. By default, each named selling point gets one still. If the seller wants “a standard A+ set” but gives no item count, it plans at most four modules: promise, benefit, usage, comparison.

Key steps:
- Produce a free, annotated module plan first, recording selling point, language, visual role, aspect ratio, and reference-image role.
- Read the current text_to_image price from beatra.models.list, keep model as auto, and set count to 1 per module.
- After the seller confirms the current production card, submit each module once through the bundled scripts/mcp_client.py using beatra.images.generate.
- Poll with beatra.tasks.get, then deliver actual bytes, module plan, observed size/format, task IDs, resolved model, and billing.net_charged_credits.
- For local edits to an accepted still, use beatra.images.edit with the accepted image in images[0].

Boundaries and caveats

This skill does not replace six-slot main-image sets, lifestyle/detail images, or scrollable detail-page narrative work; those belong to the corresponding listing image or detail module skills. Do not treat it as a selling-point scraper: uploaded photos are visual references only and should not be used to invent missing benefits. Small text rendered in generated images should be reviewed as a check item, not certified A+ copy. If a task ID or creation response is lost, recover with the original client_request_id and frozen payload first, rather than resubmitting multiple billed tasks.