Elys Listen brings local speech recognition to Unreal Engine 5.8. It captures microphone audio and transcribes speech directly on the player's Windows PC with the bundled whisper.cpp backend. Audio is not sent to an external transcription service.

Video demo| Documentation | Discord Community | Unreal Engine 5.8 only

Use the Blueprint-first API to add voice commands, dialogue input, subtitles, accessibility features, or conversational gameplay. A ready-to-use UMG interface provides microphone controls, listening status, live voice-activity feedback, completed transcripts, model selection, and language configuration. The widgets can be used as shipped or restyled for your game.

Whisper model files are downloaded separately from the editor's Elys Listen model manager. After a model is installed, transcription works locally without an internet connection.

KEY FEATURES

• Local speech-to-text powered by whisper.cpp

• Blueprint and C++ APIs

• Start, stop, mute, flush, and push-to-talk controls

• Automatic voice activity detection and manual recording modes

• Configurable transcription language

• Final text events and structured transcription results

• Runtime status, subtitle, transcript history, diagnostics, and control-panel widgets

• Editor model manager with several model sizes

• Required engine plugin: AudioCapture (built into Unreal Engine and enabled automatically by Elys Listen); no external marketplace plugin dependency.

• No cloud transcription service or required API key

IMPORTANT SCOPE

• Unreal Engine 5.8 only

• Windows 64-bit only

• Whisper model weights are not bundled with the plugin archive

• Internet access is required only when downloading a model through the editor

• A working microphone and the platform's microphone permission are required for live capture