Hello Kitty AI Voice Assistant
A voice-activated AI assistant that responds to the wake word "Hello Kitty". It listens through the microphone, converts speech to text, sends it to ChatGPT or Gemini for a response, and speaks the answer back. Supports conversation memory, customizable wake word, adjustable voice speed, and switchable AI providers.
2
AI providers, swappable
Custom
wake word and voice
Local
processing by default
The problem
Most voice assistants are locked to one provider, one voice, and one wake word. We wanted to find out what it takes to build a genuinely personal one: your wake word, your choice of model, running on your own machine.
What we did
We built a full voice loop from scratch: always-listening wake word detection, speech to text, a swappable AI provider layer, then text back to speech. The provider abstraction meant switching between ChatGPT and Gemini became a config change rather than a rewrite.
- Built a provider layer so the assistant can switch between OpenAI and Gemini without touching the voice pipeline
- Added conversation memory so it follows context across turns instead of treating every question as the first
- Made the wake word, voice speed, and model user-configurable rather than hardcoded
- Kept the whole loop running locally, so audio is not shipped to a third party unnecessarily
This project became the foundation for how we build client chatbots now: never hardcode the model provider, always leave room to switch when pricing or quality changes.
Next Project
ORBIT Web Platform →Start Building with
Orbit today
Tell us about your project, get expert guidance,
and bring your vision to life.