WebAI StudioDictate product names the recognizer has never heard — then boost them with contextual biasing and watch the diff.
Contextual biasing tunes the on-device recognizer to your app's vocabulary at runtime — no custom model training, and your domain terms stay on the device.
Boost 0–10. Higher values make the recognizer favor these spellings. Phrases apply only to the "with biasing" recording.
56 demos, all running on-device.
Say "she sells seashells" as fast as you can — the on-device recognizer scores how much of it survived.
Speak into the mic: on-device captions appear instantly, and a second track renders them live in another language.
Say "table for four next Friday at seven, outside" and watch the booking form fill itself — speech in, structured data out.