Version 1.6

Experimental “diarize” feature which detects speaker based on the difference between left/right channels.
The feature is not exposed in the desktop GUI app. Available programmatically, and in the command-line examples.

Version 1.5

Performance improvements

Version 1.4

Fixed a warning isMp3Decoder when transcribing *.wav files
Implemented “Place output file to the input folder” checkbox in the desktop app

Version 1.3

A workaround for the Microsoft’s bug in their MP3 decoder MFT

Version 1.2

Implemented a flag to select between versions of the compute shaders.
Win32 demo now has an “advanced…” button on the model loading screen to change the values.

Version 1.1

Fixed a minor bug in the in-app performance profiler.

Version 1.0

Initial published version