Runtime Speech Recognizer (Real-Time, Offline, AI) is a code plugin for Unreal Engine that converts player speech to text in real time, running fully offline. It supports Unreal Engine 4.27 through 5.8 and is built on OpenAI Whisper technology.
Runtime Speech Recognizer is an Unreal Engine plugin for adding voice input to games and applications without an internet connection. It is built on OpenAI Whisper technology and converts speech to text as players talk, which suits voice-driven menus, hands-free gameplay commands and assistive features. The plugin works with any audio input source, so developers can connect it to their own microphone pipeline. Its listed Unreal target platforms cover Windows, Mac, Linux, iOS, Android, Xbox One, PS4, Nintendo Switch and several VR headsets.
Specifications
| Seller | Georgy Dev |
|---|---|
| Type | Plugin |
| Category | Engine Tools |
| Engines | Unreal Engine 4.27–5.8 |
| Distribution | Code plugin |
| Formats | Unreal Engine |
| File types | .uasset, .uproject, .uplugin |
| Unreal target platforms | Android, Gear VR, HTML5, HoloLens 2, Linux, Mac, Nintendo Switch, Oculus, PS4, SteamVR / HTC Vive, Windows, Xbox One, iOS |
| Licenses | Personal, Professional |
| Published | 2026-10-28 |
Key features
- Real-time speech-to-text recognition that runs fully offline
- Vulkan GPU acceleration on Windows and Linux, Metal on Mac and iOS
- Multilingual models supporting 100 languages with automatic detection
- Model sizes from 75 MB to 2.9 GB selectable in settings
- Automatic language model download inside the Unreal Editor
- Optional translation of recognized speech to English
- Voice activity and command detection with noise filtering
- No static libraries or external dependencies
What the Plugin Includes and How It Works
The plugin is distributed as a code plugin for Unreal Engine, with .uplugin, .uproject and .uasset files included. It recognizes speech in real time and runs entirely on the local machine, so no internet connection is needed during play.
Recognition uses GPU acceleration where available: Vulkan on Windows and Linux, and Metal on Mac and iOS. On other platforms the plugin uses CPU processing with optimized intrinsics. Voice activity and command detection help filter noise and pick out spoken commands.
The plugin offers English-only and multilingual models in sizes from 75 MB to 2.9 GB. Language models download automatically in the Editor, and recognized speech can optionally be translated to English.
Engine Compatibility and Language Support
Runtime Speech Recognizer supports Unreal Engine 4.27, 5.0, 5.1, 5.2, 5.3, 5.4, 5.5, 5.6, 5.7 and 5.8. Its multilingual models recognize 100 languages, including English, Chinese, German, Spanish, Japanese and Arabic, with automatic language detection.
Settings for model size and language are exposed for easy selection. The plugin has no static libraries or external dependencies. For faster recognition on Android and Meta Quest, the creator points to an optional Vosk-based provider.
Use cases
- Voice-driven game menus
- Hands-free gameplay commands
- Assistive accessibility features
- Multilingual voice interaction
- VR voice control
Runtime Speech Recognizer (Real-Time, Offline, AI) FAQ
Which Unreal Engine versions does Runtime Speech Recognizer support?
The plugin supports Unreal Engine 4.27, 5.0, 5.1, 5.2, 5.3, 5.4, 5.5, 5.6, 5.7 and 5.8.
Does Runtime Speech Recognizer require an internet connection?
No. Recognition runs locally, so no internet connection is required during operation. Language models are downloaded automatically in the Editor.
How many languages does the plugin recognize?
The multilingual model supports 100 languages with automatic language detection. English-only models are also available.
Which file types are included with the plugin?
The plugin includes .uasset, .uproject and .uplugin files and is distributed as a code plugin for Unreal Engine.

