Documentation
Technical notes for teams building language products.
Source-backed guides to platform architecture, streaming speech recognition, accessible speech tools and structured book digitization.
Follow web, mobile and B2B requests through the reverse proxy and Java backend to Redis caching, Triton inference and CouchDB-backed usage services. The note also maps protected eBook licensing and delivery. Implement the WebSocket flow from the start control message through binary audio streaming, server-side voice activity detection and speech-state events to interim and final transcripts. See how the Android app combines Indian-language and English speech synthesis, translation, photographed-document recognition and read-aloud output—including text-to-speech support for everyday communication. Review the production workflow for Indian-language literature: 300 DPI scanning, OCR, human correction, structural markup, standards-compliant EPUB generation and cross-platform quality assurance.Inside the Bhashini.ai platform architecture
Streaming STT: lifecycle, VAD and transcripts
Bhashini.ai LangAssist app: speech, translation and OCR in one workflow
From printed books to structured EPUB