No standalone ASR? I am looking for an ASR as a Service for audio/video files where I can use my data to train my own models. AFAIK nobody's offering that yet.