SpeakoFlow ships a fully offline voice assistant with local Whisper, llama.cpp, and TTS

MoodOdd9657 · reddit · 2026-07-24

SpeakoFlow is a fully offline desktop voice assistant

The developer is building SpeakoFlow, a privacy-first voice assistant that runs entirely on the user’s machine.

Everything stays local by default: dictation, speech-to-text, and the built-in LLM can run without API keys or cloud calls.
Core features: hotkey dictation into any app, an OpenAI-compatible chat panel, optional screen vision, local TTS with Kokoro, and optional on-device memory and web search.
Stack: Tauri 2, whisper.cpp / Parakeet, Silero VAD, llama.cpp, and Kokoro TTS, with GPU acceleration on Metal/Vulkan.
Platforms: Windows, macOS, and Linux.
Tradeoffs: Windows builds are unsigned for now, so SmartScreen prompts are expected; local LLM quality depends on the model you choose.

The repo is public, and the author is looking for feedback on low-latency local models and Rust contributors.

More from Multimodal