Bulk Optimize
Scan the Media Library and process eligible images through the safe batch queue.
Bulk Optimize processes existing Media Library attachments in small batches.
Typical workflow
- Open Bulk Optimize.
- Scan the library.
- Start the queue.
- Pause, resume, or safely cancel unstarted work as needed.
- Review scanned, optimized, skipped, failed, remaining, batch-size, elapsed-time, and total-saved feedback.
- Retry eligible failures, clear terminal failures, or use Scan all images for a deliberate full rescan.
Eligible files
The queue uses the same source eligibility rules as upload automation and manual actions. JPEG, PNG, and WebP sources are supported. SVG, GIF, unreadable files, excluded MIME types, and files outside size or pattern rules are skipped.
Batch safety
The queue works in AJAX batches so large libraries do not require one long-running request. Default batch size is controlled by the VACIMG_QUEUE_BATCH_SIZE constant.
Queue state and the last processed attachment are stored locally with prefixed WordPress options. Resuming returns only unstarted work to pending state; it does not change originals or backups.
Generate Missing Modern Formats
Use Generate Missing Modern Formats on the same screen to prepare existing media for Automatic Modern Image Delivery. The readiness scan is non-destructive and reports scanned attachments, delivery-ready and partial records, missing or stale derivative records, and estimated derivative operations.
The migration processes small batches and can be paused, resumed, cancelled, or restarted after a refresh. It preserves valid derivatives, repairs stale registry records, skips excluded/SVG/animated GIF/unreadable files, and never converts during anonymous frontend requests. Re-running the migration is safe and idempotent.
Delivery Diagnostics
The Delivery Diagnostics panel accepts a Media Library attachment ID and reports its source type and dimensions, registered sizes, relative uploads path, delivery settings, derivative state for each size, fallback reason, stale metadata, generation timestamp, and queue state. It reads the same derivative registry used by frontend delivery, so a diagnostic result and the actual selected candidate cannot diverge.