AudioStreamSpeechWhisper

OpenAI Whisper is a robust offline speech recognition system which runs locally on user device and provides transcription, translation to English, and language recognition of user speech either using a microphone, or by analysing an audio file.

This asset contains native platforms libraries of Whisper, together with user friendly Editor interface.

This asset also features Voice Activity Detection (VAD) system which is able to adapt to incoming audio and determine when to invoke Whisper recognition automatically - suitable for open mic like scenarios; it's still possible to invoke Whisper manually

Currently WIndows (x86|x64), macOS (both Intel and Apple Silicon), Linux (x64), Android/ChromeOS, and iOS are supported.

It is currently optimized primarily for Apple Silicon (M1 equivalent or better) - but runs on standalones with higher CPU usage.

For demos please see:

Windows x64 | macOS | Linux (x64) | Android/ChromeOS

Latest documentation is also available here

For support please use Unity forums