WhisperRealtime

Real-time speech-to-text transcription and alignment plugin for UE5 with multi-language support, based on OpenAI’s Whisper model. No python or dedicated server required.

Product Details

Overview

Simple

Just add one componet to your Blueprint, and you are ready to go.
No Python or dedicated server required.

Designed for best performance

Performance-optimised AI inference is executed in the thread pool to avoid blocking the game threads as much as possible.

Features

・Real-time transcription from microphone input to text in 99 languages
・Real-time translation from microphone input to English text
・Real-time alighment from microphone input to user-specified text

※To use GPU acceleration with CUDA, a supported NVIDIA GPU is required and the versions of CUDA and cuDNN listed in the manual are required to be installed.

Free Demo

Demo projects in EXE format are available for free.
Check the performance and recognition accuracy.

Download Demo↗

Download

UE Marketplace↗

Manual

Manual↗

Technical details

This blog post details the technology used in this plugin.
You can see some of the optimisations that have been made and how they have been implemented on the UE.

Blog (Japanese)↗

Video tutorials