whisper.unity

Why we included this project

Unity developers who want speech recognition without sending audio to a cloud service will find this a good fit. It's a C# binding layer over whisper.cpp, so OpenAI's Whisper models run inside a Unity project, with sample scenes and microphone input already set up. That gives you multilingual speech-to-text and language translation (German speech to English text, for example) in games, VR/AR apps, or kiosks, all offline. It runs on Windows, macOS, Linux, iOS, Android, and visionOS, with optional Vulkan or Metal acceleration. The repo ships with the tiny model so you can test immediately, and you can drop in larger weights when accuracy matters more than speed.