the $20 Smart Home: Build an AI Video Doorbell with ESP32-CAM — Live Video, Face Recognition, Two-Way Audio, and Home Assistant Integration No Subscriptions Cloud

Somewhere on your street right now, a doorbell camera is uploading video of a delivery driver to a data center hundreds of miles away - and the homeowner pays monthly for the privilege. Stop paying. Start building. Across 15 hands-on chapters, you'll build a complete AI video doorbell around a $7 camera board and about $13 more in parts - every component itemized, every assumption stated out loud. Your doorbell will: - Stream live video to any browser or phone on your network - Detect faces on-device with a neural network drawing live bounding boxes on your stream - Recognize your household by name - "Amara is at the door" - via a 60-line recognition server you run on any spare computer - Ring your phone in about a second, notification led by a snapshot and the visitor's name - Speak and listen - synthesized chime, live listen-in audio, and quick replies you trigger from anywhere - Join your smart home - porch light after dark, spoken announcements, automatic snapshot archiving - Power itself permanently - from your existing doorbell transformer, a battery, or the sun - Defend and update itself - authentication everywhere, pinned certificates, VPN-only remote access, over-the-air updates >This is not a copy-paste tutorial. All 40+ code listings are explained line by line and tested against the current toolchain (versions pinned in print - no dead APIs, no "this compiled in 2019" archaeology). Every wiring decision is justified against the board's real constraints, including the pin conflicts, boot-mode traps, and power gremlins that kill most camera-board projects. Honest limitations are stated before you hit them, safety protocols are taught where they matter, and a free companion download provides every line of source code, organized by chapter. Inside you'll learn: - non-blocking firmware patterns and dual-core architecture- camera drivers and streaming servers- soldering and weatherproofing that lasts- digital audio on a shared bus- how neural networks actually see, and how to budget them into kilobytes- enrollment, thresholds, and biometric privacy done respectfully- home automation that speaks your visitors' names- deep sleep and solar power arithmetic- a real threat model, and security proportionate to it- and the upgrade path to newer silicon. Who it's for: makers, tinkerers, developers curious about embedded systems and edge AI, smart-home enthusiasts who want cameras that answer to them - anyone who can read code with guidance, follow "connect this pin to that pin," and spend one evening learning to solder. You'll finish with a working doorbell on your front door - and with the embedded-systems, networking, AI, and security skills to build whatever comes next. The parts are the tuition. The builder is the product.