The private voice exchange is an anonymous, decentralized network of producer nodes which provide the processing power to transcribe speech to text (STT) and text to speech (TTS) in small isloated chunks. No one node will see more than a few seconds of anonymous data and no central location exists fot this information to be grouped in any meaningful way.
User input is anonymized across the network of PriVox cloud servers which deliver bits and pieces to various producer nodes across the network. No centralized user data is retained and log files are purged hourly. While this makes it difficult to debug production systems it does ensure your information is not collected in any one place. We have no central repository of information. All our data is ephemeral.
Producer nodes may be public or private. Private producer nodes are controlled by PriVox and exist within our network. Public producers are contributors who donate computing resource to the cause. They are vetted and adhere to our privacy policy and do not attempt to violate your trust. Public producers will only see a snippet of your data here and there if you allow them to, by opting-in on your profile page. By default your data is processed by PriVox owned producer nodes but we encourage you to allow public nodes to contribute. It keps our costs down and helps open source and educational institutions improve their model training data.
See also the PVX API page, How It Works and Project Goals.