Action follows the audit
The workflow ends with usable replacement files, not only a score or a list of generic recommendations.
Scan one public webpage, select heavy image candidates, optimize them locally and download a replacement bundle.
Find heavy image candidates, choose what matters, then optimize locally.
Private networks and metadata endpoints are blocked. Scripts are never executed.
Scan one public page, inspect discoverable candidates, select the assets worth changing, then compress and package them in the same local engine used for uploads.
The workflow ends with usable replacement files, not only a score or a list of generic recommendations.
The page states what static HTML discovery misses and never claims full-site, rendered-page or script-executed coverage.
The webpage URL and selected public asset URLs must be fetched over the network. Compression, previews, settings and ZIP creation run locally after import.
Enter one public HTTP or HTTPS webpage. The server validates the destination and fetches a bounded HTML response without credentials.
Review candidate sources, declared dimensions, formats and heuristic issue labels. Select up to 20 supported images.
Fetch selected assets three at a time, optimize them in browser workers and download individual replacements or a mapped ZIP.
No. It fetches one public HTML document and does not follow internal links. Run another scan for another page.
It discovers image URLs present in fetched HTML, including image, lazy-image, srcset, picture, social metadata and image-preload sources. It does not execute JavaScript, render the page or discover CSS background images.
After you select candidates, supported files are imported in a bounded queue and passed to browser workers. The ZIP can include replacement-map.json so source URLs can be matched to output filenames.