Show HN: Ved AI Voice Assistant
TL;DR
Ved is an open, offline voice assistant for the home, positioned as a self-hosted Jarvis-style setup for consumer hardware with around 12 GB of VRAM. The current pitch centers on local execution, multimodal use and keeping system load low instead of routing household voice control through cloud services or closed smart speakers. Planned work includes a server-client architecture, cheap microphone and speaker endpoints for multiple rooms, and secure free remote access beyond the local network.
Nauti's Take
Ved looks like a maker project, but that is often where useful alternatives to big voice assistants begin. The roadmap makes sense: centralize the heavy compute, connect rooms through cheap endpoints, and avoid cloud dependency unless it is actually needed.
The catch is multiroom audio. If wake-word detection, response time and echo handling are weak, a local Jarvis can feel worse than an aging smart speaker.
Briefingshow
Ved targets a real gap: people want voice assistants at home without turning every room into a cloud microphone. If local LLMs, wake-word handling, audio routing and smart-home control can run reliably on ordinary hardware, it becomes a credible alternative to Alexa-style ecosystems. The hard part is not the demo, but latency, reliability and secure networking.
Sources