Shivaani ASR
True cache-aware streaming, four lookaheads.
This page runs Shivaani Live — a genuinely incremental streaming decoder, unlike the offline model on the main page. Pick a lookahead below; lower lookahead means lower latency but a noticeably higher word-error rate, since the model has less future context to disambiguate each word.
Live microphone
About this model
Shivaani Live is a cache-aware streaming FastConformer with a CTC decoding head. Each audio block is decoded exactly once and never re-processed — real word-error-rate/latency numbers, not a redecode-a-growing-buffer approximation. Word-level timestamps aren't exposed by this checkpoint, so file download/diarization stay on the offline page.
WebSocket protocol: ws://<host>/ws/live?api_key=...&lookahead=480&source_lang=hin_Deva&target_lang=eng_Latn
— send 16kHz mono PCM16 frames, receive {type: partial|final|lookahead_changed, text, translation, ...}.
translation is only present on final events, and only when both language
params were set. Switch lookahead mid-connection with {"type":"set_lookahead","ms":80}
(language pair is fixed for the life of the connection — reconnect to change it). Full API reference
(including how to get your own key) is in the developer docs.