Open-source constructor for personal AI assistants with voice control, persona system, and full local privacy. Choose character, voice, and brain — run it on your hardware.
The era of smart speakers is evolving. Take back control of your AI assistant.
Full voice pipeline — wake word, VAD, speech-to-text, text-to-speech — all running locally on your device. No audio ever leaves your machine.
Every component is replaceable. Swap wake word engines, STT models, TTS voices, and AI backends independently. Your assistant, your rules.
Like creating a character in an RPG — pick personality, voice, skills, and appearance. Six built-in personas or create your own.
All audio processing happens locally. The only external connection is the AI text API — and even that is optional with Ollama.
Everything you need for a personal AI assistant that respects your privacy
OpenWakeWord detection — each persona has its own activation phrase. "Hey Hestia", "Hey Argus", or custom.
Fast local transcription with excellent multilingual support. Optimized for Apple Silicon and modern CPUs.
Natural-sounding voices with per-persona configuration. Adjust speed, tone, and personality.
Claude, OpenAI, Ollama, Gonka, or OpenClaw. Switch providers without changing anything else.
React Native companion app — remote control, push notifications, status dashboard, media control.
Home Assistant integration. Control lights, temperature, scenes — all by voice.
Free remote access via Telegram. Send commands and receive notifications from anywhere.
Visual browser-based wizard guides you through persona selection, voice config, and AI backend setup.
Fully autonomous with Ollama + local LLM. Perfect for remote locations with unstable internet.
Choose a character that fits your needs — or create your own
Home Keeper
Warm, caring, patient. The goddess of the hearth, welcoming and nurturing. Perfect for family and daily tasks.
Strategic Advisor
Smart, concise, strategic. The goddess of wise counsel who always knows the right move. For planning and analytics.
Wise Mentor
Patient, thorough, educated. A sage from the Iliad. Great for education, children, and detailed explanations.
Companion
Empathetic, soft, supportive. "Wisdom" in Greek. Your thoughtful companion for reflection and planning.
Guardian
Serious, attentive, concise. The all-seeing mythological watchman. For security and home monitoring.
Custom Persona
Define personality, voice, skills, and system prompt. Package it as YAML + markdown — share with the community.
Not another CLI tool. A real visual experience — because your assistant should feel personal from the first minute.
From wake word to spoken response — entirely on your device
OpenWakeWord listens for the persona's activation phrase. Always-on, always local.
OpenWakeWordSilero VAD monitors for speech. Stops recording after 3-4 seconds of silence.
Silero VADwhisper.cpp transcribes speech to text locally. Fast on Apple Silicon.
whisper.cppText is sent to the selected AI backend with the persona's system prompt.
Claude / Ollama / OpenClawTTS converts the response to speech using the persona's configured voice. Streamed sentence-by-sentence.
Piper TTSOne command installs everything — dependencies, models, config wizard
curl -sSL https://raw.githubusercontent.com/openocto-dev/openocto/main/install.sh | bashirm https://raw.githubusercontent.com/openocto-dev/openocto/main/install.ps1 | iexopenocto startAlready cloned the repo? Run ./install.sh instead.
Runs 24/7 on a Mac Mini. Silent. ~$10/year in electricity. Always listening, always local, always yours.
Building the future of personal AI assistants, one phase at a time
Updates, guides, and insights from the OpenOcto project
Why we're building an open-source voice assistant constructor with a persona system, local privacy, and no vendor lock-in.
A deep dive into the local voice pipeline: OpenWakeWord, Silero VAD, whisper.cpp, and Piper TTS — no audio leaves your device.
How personas work under the hood — YAML configs, system prompts, voice settings, and how to create your own custom persona.
OpenOcto is built by the community, for the community
Star the repo, report issues, submit PRs
Russian-speaking community channel
Questions, partnerships, commercial licensing
info@openocto.dev
Built with love by Rocket Dev 🚀