Bulk AI metadata batches
Run template-based AI metadata across many images with preflight checks, progress tracking, and bulk human review.
Orchestration, not auto-publish
Bulk batches call the same single-image AI metadata service used elsewhere — they do not bypass human review or auto-approve output.
Each eligible image is sent to the configured AI provider as a temporary analysis copy; originals remain in private R2 storage.
Drafts must be reviewed and explicitly approved before they become approved metadata for exports or integrations.
Templates and languages
Choose a fixed template: SEO, accessibility, or e-commerce. Templates define prompt version and output fields — there is no custom raw prompt field in the dashboard.
Output language can be English or Urdu when supported by the template, independent of the interface locale.
Preflight resolves ownership, eligibility, existing drafts, rate limits, and monthly AI allowance before a batch starts.
Progress and review
Batch progress shows real counters: queued, running, draft, failed, cancelled, stale, and skipped items.
Active batches can be cancelled; failed items can be retried when policy allows.
When drafts are ready, open bulk review to inspect alt text, title, quality scores, and duplicate flags — then approve or reject selected rows with confirmation.
Limits
Up to 50 images per batch and a small number of concurrently active batches per workspace.
Monthly AI allowances and per-image/per-project daily generation limits still apply.
Workflow automation can also trigger metadata batches via the generate_metadata_batch action.