OpenVoiceOS

OpenVoiceOS is a voice assistant platform for people building smart speakers or adding voice control to their own devices. Its core is a fork of MycroftAI/mycroft-core, and most classic Mycroft skills remain compatible.

• Voice tasks: Skills provide smart home control, music, podcasts and radio playback, answers to questions, timers and reminders.
• Custom assistants: Builders can choose wake words, adapt the interface and extend the assistant with community skills or their own code.
• LLM conversations: ovos-persona handles input that skills can't answer and can connect to a local OpenAI-compatible LLM server or use an OpenAI plugin.
• Local speech processing: Vosk and Mozilla DeepSpeech provide offline speech recognition, while Mimic3 provides local text-to-speech; cloud alternatives include Google TTS and Amazon Polly.

The core runs locally without a backend internet server, but default speech recognition requires internet access. Local speech plugins can keep recognition and speech generation on your hardware. Individual skills and other plugins may still need a connection, and cloud speech services process requests externally.

OpenVoiceOS is free and open source under Apache 2.0, which permits personal and commercial use. It runs on Linux desktops and laptops, Raspberry Pi, Mycroft Mark I and Mark II devices, and in Docker. A dedicated Linux image supports embedded devices, including headless hardware and small touchscreens.