Show HN: Serverless speech recognition with WebAssembly

In this thread:

The Zork game, written in WebAssembly, using serverless (offline) speech recognition by using the Kaldi Speech Recognition Toolkit. Kaldi is written in C++ which then (i guess) is compiled into WebAssembly via Emscripten.

https://github.com/kaldi-asr/kaldi

Very very interesting.

Additional comment: When I saw the Amazon Alexa API (i was at a small meeting which included defining your own "utterances", i was dissapointed that the Alexa required internet connection to perform the speech recognition, even moreso when the interaction with it was not so sophisticated (after all, it is geared to understanding predefined commands).

Alexa's recognition is made more difficult by the fact that you talk to it from a distance. Still, though, you're right that only handling predefined commands should make it much easier.

Deadly. Have always hated the privacy implications of server-side speech processing.

Super super cool. What data did you use to train the acoustic model?

see https://groups.google.com/d/msg/kaldi-developers/jakcCUK5lKY...

It seems to be a heavily compressed version of the ASpIRE challenge example model (https://github.com/kaldi-asr/kaldi/tree/master/egs/aspire/).