Show HN: TuxWhisper – Offline push-to-talk dictation for Linux, works on Wayland (github.com)

TuxWhisper has emerged as a groundbreaking offline dictation tool for Linux, allowing users to seamlessly dictate text through a simple hotkey interface. By leveraging the faster-whisper engine, speech recognition occurs entirely on-device, ensuring user privacy by keeping data off the cloud. This functionality is particularly significant for Linux users relying on Wayland, as it fills a gap where many typing tools fail to operate. The tool boasts a quick response time, processing commands within about half a second when using an NVIDIA GPU, offering efficiency that can enhance productivity across various applications. What sets TuxWhisper apart is its versatility and ability to refine the transcription process with an optional local LLM cleanup using Ollama, which improves punctuation and clarity by removing filler words. Additionally, it supports various keyboard layouts and integrates smoothly with any application that accepts text pasting. TuxWhisper also allows users to build a voice dataset for future fine-tuning of speech models, making it not just a dictation tool but also a platform for advancing speech recognition technologies. This innovation presents exciting opportunities for developers and researchers in the AI/ML community, emphasizing local processing capabilities and diversified usage scenarios.