Provides streaming API for the best user experience (unlike popular speech-recognition python packages).Portable per-language models are only 50Mb each, but there are much bigger server models available.Works offline, even on lightweight devices - Raspberry Pi, Android, iOS.
