Single-item AI listing solved "how one SKU gets approved." What still burned the day was "how a pile of new products gets through together." Collect, jump tabs, patch images, wait for review, delete on reject — Batch Listing folds that path into one workbench.

Open New Listing → Batch Listing. Manual fill / spreadsheet import on one side, a sticky Collection Pool on the other, quota bar always on top. Cap is 20 rows per batch — enough for a real drop, without freezing the page.

What this feature does for you

Collection Pool → batch in one click: the browser extension collects a 1688 product page into the pool (images, attributes, official weight/dims resolved server-side). Select rows → Add to batch / Import to editor — they leave the pool and land in the editor, no copy-paste across tabs. Amazon and other sources can follow the same pool → batch path.
Reference / main / supporting slots stay separate: reference images are for AI to see the physical product (up to 8) and are not forced in as Takealot main images; main is 1 slot; a supporting image can be promoted to main. Drag a reference thumbnail into main/supporting and the system re-uploads via our relay — no download-then-reupload. Spreadsheet import carries text only; images always go through the three slots.
AI fill + image gen, with row-level stage dots: with AI images on, each row runs classify → fill → generate → submit. The batch progress card marks green / amber pulse / red by stage so you see where it stuck. Per row you set a target image count (2–10, never below the category floor). Need sharper detail? Toggle Higher quality (2048 + high, billed at 2× points).
Image quota always visible, top-up anytime: the bar shows free quota (Standard 10 / Advanced 25 / Professional 50 images per subscription period; trial = 0 — generation is real third-party cost) + purchased points + Recharge. Dual entry: Settings → Subscription & quota, and beside the batch AI toggle. Alipay tiers credit on scan (¥15 / 30 / 50 / 100 / 300 / 500; larger tiers show a "+N images" badge). Pre-create checks report free remaining and paid balance separately — it will not silently under-charge a whole batch.
Rejected → edit & resubmit; partial batches are not dead ends: rejected rows in submission history get an orange Edit & retry — failure reason pinned at the top, full product form inside, submit creates a fresh run while the old reject stays for audit. If only some rows fail pre-submit, review accepts partial; Resubmit needs-fix pulls those rows back into the submit set, with red badges pointing at the exact fields.
Per-row sourcing / price + batch direct-mail: each editor row takes a sourcing URL (prefilled from collect, editable; 1688 / Pinduoduo / Taobao auto-detected) and an integer ZAR selling price. Batch-level Enable direct mail after approval defaults off; submit fails fast if your seller account has not enabled lead time covering 14 days. After approval the system binds source, sets price (RRP at 1.5×), and turns on direct mail by SKU — no second pass on the product page.
Variants on the same pipeline (optional): colour and other variants ride the batch / pool path; review embeds variant editing per row. Images must not repeat across variants (manual review rejects duplicates). For colour variants, AI recolour generates a dedicated set from the base images before submit.

Boundaries, plain: 20 rows per batch; extra reference images never silently become the main image; free quota rolls with the subscription period and trials get none; Higher quality bills 2× — check balance before you flip it. Missing dimensions / weight surface three times (import toast, amber row warning, pre-submit summary) — better you fill them than Takealot bounce with DIMENSION_MISSING.

Real seller scenarios

A morning of 1688 collecting: extension grabs a dozen detail pages into the pool, you import 15 into the editor, turn on AI images at 5 per SKU, and go purchase. Green dots are complete when you return; one review pass, whole batch to platform review.
Spreadsheet first, photos second: fill the template with titles, category intent and prices, import, then drag product photos into the reference slots. AI writes the English listing and main/supporting set from those references — cleaner than pasting image URLs into Excel.
Reject without delete: wrong category or enum → Edit & retry, failure banner still visible, fix and resubmit. The original reject stays for comparison; you do not restart from the pool.
Price and direct mail set on the way in: fill sourcing URL and selling price per row, flip the batch direct-mail switch. The night after approval, products already have source bound, price set, and 14-day direct mail on — no second click-through the next morning.

AI Batch Listing is part of 辉光ERP — the all-in-one platform for Takealot cross-border sellers in South Africa. New products need speed; a pipeline beats overtime.