Before the Mark II ships in September, local STT will be available by default!
> In order to provide an additional layer of privacy for our users, we proxy all STT requests through Mycroft's servers. This prevents Google's service from profiling Mycroft users or connecting voice recordings to their identities. Only the voice recording is sent to Google, no other identifying information is included in the request. Therefore Google's STT service does not know if an individual person is making thousands of requests, or if thousands of people are making a small number of requests each.
Well, unless Google does voiceprint analysis. But Google wouldn't do that, would they? /s
Beyond that, if I'm reading right, local STT will still require a separate STT server. It won't run on the Mark II itself, right?
> Mimic 3: Mycroft’s newer, better, privacy-focused neural text-to-speech (TTS) engine. In human terms, that means it can run completely offline and sounds great. To top it all off, it’s open source.
If "skills" are what I think they are (something like external commands, for example "Play X on Spotify"), then my understanding would be that everything but those runs offline and local-only.
But if things like `speech to text` requires internet connection and sends the data to some cloud service, then the entire value proposition of this product falls apart.
I hope that's really not the case, as that would be outright lying and false advertisement.
There are some options for that available if you're running Mycroft already: https://mycroft-ai.gitbook.io/docs/using-mycroft-ai/customiz...
But these are not what will be shipped on the Mark II...
It's very unclear and misleading to put "it can run completely offline" if you're not 100% sure it can actually run "completely offline", hardware be damned.
"*When you use our Services including the Mycroft Voice Assistant, your voice and audio commands are transmitted to our Servers for processing.*"
So it appears that STT is still cloud-based, which is a pity. That's the only thing keeping me from ordering one today.
As will a new privacy policy that better reflects what we actually do.
Note: I'm from Mycroft
I’ve seen you say that Mark II is going to be able to be run non-internet.
Will I be able to be run on my local network, even if I have to dedicate a machine to just speech to text (STT) ?
There are several ‘layers’ to a voice assistant;
Wake Word - that detects when you are speaking to the device. This is local to the device and we use PocketSphinx.
Speech to text - that detects what you say to determine Intents - we currently use a cloud service for this
Intent matching - this is done locally using our own open source software - Adapt and Padatious
Skills - Intents then match to Skills. Some Skills require internet connectivity.
Text to Speech - We use our own software called Mimic for this, it’s local to the device.