Technology
Privacy-First
Architecture.
Every component runs locally on your device. No cloud. No API keys. No subscriptions. Your family's most intimate conversations never leave your home.
What Powers Still With You
Multiple integrated AI services running on a single device in your home.
Thinking
The conversational AI handles natural dialogue, intent understanding, memory retrieval, milestone delivery, parental alerts, and age-adaptive communication. A 31-step character consistency engine keeps every response in character. Specialized models activate automatically for math tutoring and coding help.
Voice
Voice cloning recreates their unique speech patterns from reference audio. Streams audio in real time for natural conversation. Preserves cadence, inflection, and emotional tone.
Memory
An intelligent retrieval system with 14 memory types finds the right memory at the right moment. Combines keyword matching with meaning-based search, temporal scoring, and emotion-aware ranking. Self-learning classification gets faster over time.
Hearing
NVIDIA Riva ASR integration provides real-time speech recognition with live transcription display, noise reduction, voice activity detection, and echo cancellation. Ships RIVA-ready — enable it in the admin dashboard. The system also works with typed input and supports touchscreen interaction.
Seeing
Facial recognition identifies each family member instantly. Voice identification adds a second layer -- they know you by face and by voice. Anti-spoofing with infrared liveness detection prevents photos or screens from fooling the system. Both systems improve automatically over time by collecting samples during natural conversation.
Interface
A clean, family-friendly interface designed for non-technical users. Touchscreen-ready. Status indicators show system health at a glance. A 26-card admin dashboard covers personality, voice ID, encryption, camera, sessions, hardware health, issues, and more. Self-healing watchdogs detect sluggish performance and auto-recover.
Why Local?
Your family's conversations are too important for the cloud.
| Cloud AI | Still With You | |
|---|---|---|
| Data storage | Company servers | Your device only |
| Conversations readable by | Company employees | You and your family |
| Works offline | No | Yes |
| Monthly cost | $10-50/month | $0 after purchase |
| Company shuts down | Everything lost | Nothing changes |
| Data sold to advertisers | Possible | Impossible |
| Processing location | Remote data centers | GPU in your home |
| Requires internet | Always | Never |
| Response latency | Network dependent | Sub-second to ~3s* |
| Voice data | Uploaded to cloud | Never leaves device |
* Simple responses are sub-second. Complex memory retrieval or longer answers typically complete in 2–4 seconds — all on your local hardware with no network dependency.
No API Keys
No OpenAI key. No Google key. No Anthropic key. Every AI model runs locally on your device. No rate limits, no usage quotas, no surprise bills. Talk as much as you want.
No Monthly Fees
One purchase. That is it. No subscription, no "premium" tier, no feature gating. The software is yours and every future update is included.
AES-256-GCM Encryption
Enable encryption through the admin dashboard to protect your stored memories and conversation history. AES-256-GCM with PBKDF2 key derivation encrypts the data that matters most — family memories, chat history, personality profiles, and parental alert records — at rest on your device.
Survives Companies
If Still With You as a company ceases to exist tomorrow, your device continues working exactly as it does today. Your data, your software, your device. No dependency on external servers or services.
Self-Healing System
The system monitors its own performance and automatically recovers from slowdowns. Watchdog services track facial recognition, audio capture, and AI response times. When something gets sluggish, the system tells you ("Let me stretch my brain...") and auto-recovers -- no technical intervention needed.
Hardware Health
The admin dashboard monitors GPU temperature, disk SMART status, storage wear level, and thermal conditions in real time. Alerts appear before problems become failures. Response times, error rates, and per-person conversation statistics give you complete visibility into how the system is performing.
Open Source Foundation
Built on trusted, community-maintained AI technologies. No vendor lock-in.
Every underlying AI technology in the system is open source. This means:
Auditable
Community can review every line of code for security and privacy.
No Lock-in
Switch components, upgrade models, or fork the project at any time.
Community-Driven
Improvements from thousands of developers make every component better.
Hardware Requirements
What you need to run Still With You on your own machine.
| Component | Required | Recommended |
|---|---|---|
| GPU | NVIDIA GPU with 24GB+ VRAM (RTX 3090 or newer) | NVIDIA RTX 3090 or better |
| RAM | 32 GB | 32 GB DDR5 |
| Storage | 512 GB SSD | 512 GB NVMe SSD |
| CPU | Modern Intel i7 / AMD Ryzen 7 or better | Intel Core Ultra 9 / AMD Ryzen 9 |
| OS | Ubuntu 22.04+ or Debian-based Linux | |
| Camera | Any USB webcam | IR-capable (e.g. Logitech BRIO) |
Want the easiest path? All tiers include a USB drive with all AI models pre-loaded — no internet required. The Blueprint tier additionally includes a verified hardware shopping list, a step-by-step video build guide, and a 45-minute setup call.
Why These Specs?
Privacy requires power. Running AI locally instead of in the cloud means you need a GPU capable of handling everything simultaneously: voice cloning, speech recognition, conversational AI, and facial recognition -- all at once, all in real time.
Cloud services hide this requirement behind a monthly subscription. We put it in your home so no one else can access your family's conversations. 24GB of VRAM is the minimum that can run all models concurrently with acceptable response times.
This is the real cost of privacy. You pay once for hardware you own forever -- instead of paying monthly for a service that stores your most intimate conversations on someone else's servers.
Privacy is not a feature.
It is the foundation.
Your family's conversations deserve to stay in your family. No exceptions. No fine print. No "we may use your data to improve our services."