Show HN: Ved AI Voice Assistant
TL;DR
Ved is an open-source attempt at a fully offline, self-hosted voice assistant for the home, framed as a personal Jarvis-style setup. The developer targets everyday consumer hardware, mentioning a 12 GB VRAM GPU and a multimodal system that should avoid dragging down the main machine. Planned next steps include a server-client setup, cheap microphone and speaker endpoints on Raspberry Pi-like hardware, and secure remote access from outside the local network.
Nauti's Take
Ved is the kind of project that starts from a very practical frustration: smart homes should feel like science fiction, not like a subscription with a microphone. The real test is not whether it can produce one good demo.
The test is whether it hears reliably in a sleepy morning routine, answers fast, stays out of the way and does not become another admin chore. Local AI needs audio, permissions and updates to be treated as seriously as the model.
Briefingshow
Offline voice assistants hit a real pain point: people want smart-home convenience without a cloud box listening in the living room. Ved points to where the builder layer is moving: local models, small audio endpoints, central home servers and more control over data. The hard parts are latency, maintenance, security and whether the setup survives daily use.
Sources