Indian-English model (36 MB, cached after first load), runs 100% in your browser. No server. Hold the button, say your order, release. → Compare with Whisper
Loading model… (first load downloads ~36 MB, then cached)
Recognition latency (release → result): —
Transcript
—
Matched items
Try: “two flower pots”, “ten seven cm electric sparklers”, “gold lakshmi and five atom bomb”, “clear”.
Catalog is a demo list (see the items it maps to). Say product names in English.