Anatomy of an open source voice assistant
summary
Mycroft presents an open source voice assistant stack: a Mark 1 Raspberry Pi enclosure with a wake word that currently false-triggers, a cloud STT pipeline and a message-bus core that routes intents to skills. The intent stack uses adapt intent parser (100% Python), a neural classifier (palacios) and Pocketsphinx fallback; TTS is mimic TTS. Community-run skills live in a Github repo, installation via MSM or a Pycroft image, and there are 4,406 registered units reported.
AI-written summary · about
chapters (18)
- 0:00Introduction
- 2:00What is Minecraft AI
- 3:40Open Source AI
- 4:23Hardware overview
- 10:39Community
- 11:35Channels
- 13:36Users
- 14:50Skills
- 16:13Port
- 16:54Voice Assistant
- 20:01Message Bus
- 21:36Wake Word
- 22:42Sphinx
- 29:56Scalar
- 31:19Intent Manager
- 32:38Facial Neural Networks
- 39:21Voice Types
- 40:55Developing Skills
related talks
36mMyCroft
41m"Hey Mycroft": Getting Started with the OSS Home Assistant
45mComparison of Voice Assistant SDKs for Embedded Linux Devices
45mVoice Assistants & IoT Automation
24mOpen Source Voice AI: From ChatGPT's Infrastructure to Your App
45mMore voice, less choice: The rise of voice interfaces and the decline of open source voice
30mMindmeld: An Open-source, Deep-domain Conversational AI Toolkit
26mOpen AI Assistants