Recognissimo is a cross-platform offline speech recognition plugin.

Features:

• No internet connection required
• Supports 21+ languages and dialects and more to come
• Fast and lightweight, written in C++
• Setup everything in Editor without code
• Load languages from the web or local storage
• Automatic storage and permissions management
• Use speech data from a microphone, audio clip or Unity scene
• Includes speech recognizer, voice activity detector and voice control with regular expressions support
• Extend using simple API

Supported platforms:

• Windows (x86, x64)
• macOS (x64, ARM64)
• Linux (x64)
• Android (ARMv7, ARM64, x86, x86_64)
• iOS (ARM64, x64)
• WebGL (HTTPS required)

Supported Unity editors:

• 2021.2 and above

Supported languages and dialects:

• Arabic
• Chinese
• English
• French
• German
• Italian
• Portuguese
• Russian
• Spanish
• Catalan, Czech, Dutch, Esperanto, Farsi, Filipino, Hindi, Indian English, Japanese, Kazakh, Swedish, Turkish, Ukrainian, Vietnamese

Recognissimo uses Vosk as a speech recognition backend, so you can use any Vosk-compatible models.