Tater – Tater v80
🥔 Tater v80 — “Wake Word Wars” is here! 😏🎙️
The Tater ecosystem just leveled up from a simple tool to a full-blown laboratory for voice activation! This update turns your local AI setup into a modular powerhouse, allowing you to mix and match different wake engines across all your satellites.
What’s New:
- Multi-Engine Support: You can now run `microWakeWord`, `openWakeWord`, and the brand new `NanoWakeWord` all within the same ecosystem. The best part? You can switch between them live via firmware entities without breaking a sweat. 🎙️
- Dedicated NanoWakeWord Server: We’ve added a standalone server for NanoWakeWord! It includes Docker/NVIDIA Docker support and dedicated WebSocket streaming.
Pro-tip:* Use this to stream audio from your satellites to a separate backend, offloading the heavy processing from your main Tater app! 🛰️
- NanoWakeWord Trainer: For the true tinkerers, there is now a dedicated training flow. It features synthetic sample generation, validation handling, and seamless ONNX + PyTorch model syncing. You can build and deploy your own custom models directly into the ecosystem. 🏋️
- openWakeWord Cleanup: We’ve dialed back the noise by making diagnostic detection logging optional and disabling it by default. Your logs will finally stop screaming at you! 🔧
The Big Picture: 🧠
Tater is officially modular. Whether you want local detection, remote detection, or a custom-trained model running on a dedicated server, you can now build a highly customized, distributed voice ecosystem. We didn’t just add an engine; we started a collection!