NexTalk – Offline voice input with low latency for Linux

Summary: NexTalk is a modern voice input tool designed exclusively for Linux, offering 100% offline speech recognition with sub-20ms latency. It integrates natively with Fcitx5 and features a minimalist transparent UI for seamless desktop use.

What it does

NexTalk performs local voice inference using Sherpa-onnx, providing private, offline dictation. It connects directly to Fcitx5 via Unix sockets, ensuring compatibility with Wayland and Linux applications.

Who it's for

It targets Linux users seeking a native, private, and responsive voice input solution without relying on cloud services or workarounds.

Why it matters

NexTalk addresses the lack of integrated, high-quality voice dictation on Linux by delivering fast, private, and native offline speech recognition.