How Nepali Voice Typing Works — Web Speech API Explained
नेपाली भ्वाइस टाइपिङ कसरी काम गर्छ?
Nepali voice typing works by capturing your microphone input in the browser and sending the audio stream to the browser's Web Speech API, which uses a cloud-based language model trained on Nepali speech to return Unicode Devanagari transcription in real time.
What is the Web Speech API?
The Web Speech API is a browser interface that exposes speech recognition and text-to-speech to web applications. Chrome and Edge implement it using Google's speech models; Safari uses Apple's models. TypeNepal calls the API with language set to ne-NP for Nepali.
What happens to my audio?
Audio is streamed directly from your browser to the speech recognition service operated by your browser vendor (Google for Chrome, Apple for Safari). TypeNepal never touches the audio. The browser returns text, and TypeNepal displays that text inside the editor.
How accurate is the recognition?
Accuracy depends on background noise, microphone quality and accent. In a quiet environment with a clear voice, modern models reach 90–97% on conversational Nepali. Domain-specific vocabulary, names and technical terms typically need light editing after dictation.
Does voice typing work without internet?
No. The Web Speech API uses cloud language models, so an internet connection is required for live transcription. Once text is in the editor it can be saved, copied or exported entirely offline.