Confirmed, we don’t have a WiFi module on board.
I’ve responded about the data in Pure here: Reviews of Mudita Pure, so I’ll just focus on the speech-to-text functionality.
At first, let me explain to you how such a functionality works (in most cases):
- you speak to the phone, the voice is recorded
- the audio data is being processed (offline or online)
- the result is speech converted to text
The first point can be concerning - we have to record the words that you want to convert into text. It might be a privacy issue for some of the users, but alright, that’s not too bad. And then the second point comes in - data processing.
To be able to do this offline (without using any 3rd party services), the processing would require pretty huge computing power for analyzing the audio samples and using/training the neural network. Please consider also that every new language will require a new dataset, which is also occupied with increased power usage.
Of course, it could be done via 3rd party service such as Google Speech-to-text API, but we don’t use the mobile data inside our OS and we don’t want to send private data to Google.
We can’t ignore the fact, that those companies had trained the systems for years to have such accuracy.
Hope that sheds some light on this complicated functionality and explains why we don’t have this right now.