To be clear, Mycroft does not by default run completely locally or offline. It is possible to do so, but requires you to run a range of services yourself which is beyond the average user. The speech recognition is still cloud based, as it requires more computing resources than the vast majority of people have or are willing to allocate, at least to process the audio in a reasonable time frame and with the accuracy required. Our more robotic sounding voices like British Male are generated on device, but the premium voices are also cloud based. Again they require significant compute power to synthesize the audio quickly enough for natural interaction.

Hi there, I wanted to clarify a few things.

Everything in Mycroft is open source and available on our Github except some of the premium pre-trained voice models. The code running on your device is not obfuscated in any way so you can verify that it is the exact code available in the open source repositories.

By default it does require an account on our website to manage devices and access services. You can self-host the backend or an alternative like our community built Personal Backend as they are both also open source.

This does require you to run your own supporting services like speech recognition. There are some self-hosted options like DeepSpeech or Kaldi, however in our experience these don't yet provide sufficient speed and accuracy to be usable for a voice assistant. We are continuing to work with Mozilla on their DeepSpeech project toward this goal though.

Mycroft is owned by Mycroft AI Inc. We have leased office space from Middlebrow LLC.

Unlike other assistants, by default we don't retain your data. We only retain data for those who explicitly opt-in to anonymously share it for the improvement of the service. We understand that people have different security requirements and we try to meet those needs where ever possible.