🎙️ Whisper WebGPU POC — fully offline

OpenAI Whisper (base, multilingual incl. Tamil) running on your GPU via WebGPU. No server. Records while held, transcribes on release. → Compare with Vosk

Loading model… (first load downloads the model, then cached)

Language: EN TA

Transcription latency (release → text): —

Transcript

—

Matched items

Try (EN): “two flower pots”, “gold lakshmi and five atom bomb”. Try (TA): “இரண்டு பூச்சட்டி”, “gold lakshmi ஐந்து”.