[phillipseamore]

This is something I've thought about as well, though not for security reasons per se, but to lower the barrier to entry and something used as a solution in the beginning but to be discarded as trust is built.

All the parts are known, the same techniques used to put bunny ears on a head can be used to blur or replace faces and it's possible to animate avatars on static backgrounds with the physical movements captured in video. Audio can easily be manipulated to mask a voice (but would need to be randomized to minimize reverting it). Speech-to-text and text-to-speech could be utilized to mask a voice as well, though it introduces some latency and would mask various vocal cues a therapist listens for.

This could also be used to minimize prejudice with a therapist unable to know age or gender for example. But those can be important aspects of analysis as well.

So I've got no ready made solutions for you, but good luck.

[sampkola]

Thank you for the response. You hit the nail on most aspects that I had in my mind. I agree that whatever the solution (realistic face transfer or any other way to retain facial anonymity by turning off the camera like the other suggestion) it needs to lower the barrier of entry.