March 16, 2026
Ring, ring—your AI is calling
Show HN: Hecate – Call an AI from Signal
You can now video‑call an AI on Signal — and everyone says voice is the real headline
TLDR: Hecate lets you place real voice/video calls to an AI inside Signal, pushing bots from chat windows to your contacts list. Commenters say the headline is the voice/video feature, not texting; Linux users cheer, Mac users gripe, and privacy/setup quirks spark jokes and raised eyebrows about auto‑answering and CPU burn.
Meet Hecate, the DIY project that lets you literally call an AI through Signal, the privacy‑focused chat app. The maker says it shines on Linux but struggles on Mac, and the setup is a bit wild: you’ll need Signal on your phone and on an emulator, a second phone number, and even OBS to fake a camera for video. The payoff? A talking avatar that picks up your call and chats back, all inside encrypted Signal calls.
The early crowd zeroed in on the branding drama. One top reaction begged the title to say what’s actually exciting: voice and video. Texting a bot? “Meh.” Video‑calling one? Now we’re listening. That set the tone: Linux folks flexed (“works great!”), Mac users groaned (“works poorly…”), and everyone joked about asking the bot movie‑night questions like “Which Pirates of the Caribbean is best?” while watching their CPU melt. Someone side‑eyed the irony that a privacy project pipes through an emulator that auto‑answers unknown calls until that gets fixed, even as Signal’s end‑to‑end encryption keeps the line private (signal.org).
Beyond the memes, people see a bigger shift: this turns AI from a chat window into a phone contact. If “FaceTiming your bot” becomes normal, expect a lot more talking to machines—and even more debates about convenience versus creepiness.
Key Points
- •Hecate enables voice and video calls to an AI using Signal, with strong Linux support and weaker Mac support.
- •Setup requires Signal on both a phone and an emulator, plus registration with a second phone number.
- •Voice calls run via a loop process; video calls are implemented using OBS Studio’s virtual camera due to high CPU with Docker + Chrome.
- •Configuration via .env supports multiple STT (e.g., whisper-large-v3-turbo) and LLMs (e.g., llama3-3-70b, deepseek-r1-0528), along with voice and avatar selections.
- •Security uses Signal safety numbers; the emulator currently answers all calls without filtering, but the AI has no memory between calls. The project is MIT-licensed.