Minimal
Runs everything basic
Home server, subdomain apps, cloud AI, light local AI
CPU
4 cores — Intel i5 (8th gen+), AMD Ryzen 5, or Apple Silicon M1.
Memory
8 GB — enough for the home server + browser + Ollama on a small model (TinyLlama, Phi).
Storage
256 GB SSD — OS + VP3 + a few brain folders. Add external storage for media archives.
GPU
Integrated / none — image gen + Whisper fall back to CPU. Slower, still works.
Network
25 Mbps up — enough for a handful of concurrent visitors. Cloudflare Tunnel handles SSL + DDoS.
OS
macOS 12+ / Windows 10 / Ubuntu 20.04 / Pi 4. Single Electron binary, no Docker.
Browser
Chromium or Safari for the operator dashboard at localhost:3077.
Optimal
Full local-AI workstation
Llama-13B, SDXL, Whisper-Large, Remotion render, all local
CPU
8+ cores — Intel i7/i9 (12th gen+), AMD Ryzen 7/9, or Apple Silicon M2 Pro / M3.
Memory
32 GB — Llama-13B + Stable Diffusion XL + browser + render queue in parallel.
Storage
1 TB NVMe SSD — model weights, video renders, family-brain attachments, never running out.
GPU
NVIDIA RTX 3060 (12 GB)+ or Apple Silicon (M-series). SDXL + Whisper-Large run at real-time speeds.
Network
100 Mbps+ symmetric — comfortable for live streaming, render uploads, busy storefronts.
OS
macOS 14+ / Windows 11 / Ubuntu 22.04+ / VP3 Cauldron. Cauldron is the purpose-built appliance.
Browser
Latest Chromium or Safari — WebGPU + Web Audio + MediaRecorder all stable.