Is SpeechKit speaker dependent or independent? If speaker dependent, what is involved in training the speech engine?

Speaker Independent Speech Recognition

Speaker dependency is a feature of the recognition mode of a speech recognition engine not of the Chant components.

Command and Grammar modes of most engines today are speaker independent. The Chant components provide a simple way to invoke a word training capability to add pronunciations to the user lexicons for proper nouns that the recognizers may not handle well in these modes depending on the speaker.

Dictation mode in most engines today works better if the speaker does training. The speaker or end-user uses a tool provide by the recognizer (e.g., ViaVoice, MS Dictate, etc.)

Chant is licensed to bundle Microsoft royalty free engines with SpeechKit. This means you can develop and deploy royalty free apps with Chant and Microsoft technology.