In one of my projects we are recording voice from the user and converting it to text using a speech to text engine. While we provide support for pocketsphinx and kaldi this is simply not enough for some users, which is why we allow the use of online speech to text conversions like the google asr aswell. However we would like to provide this functionality with a minimum loss of privacy. This is why we search for a tool to anonymize the voice recordings before we send them out:
Requirements for the software:
open source
not easily invertible even when the source code is known (open source)
has to be fast (our application is realtime)
The speech to text conversion still has to work properly after anonymization
we are using python, so should be as simple as possible to call from there would be a nice to have but is not required
I did not really find an application or library that fits our requirements, so I hope someone here knows one :)
Thanks in advance