TTS watermarks
Preventing deep fakes[edit | edit source]
Based on Erogol's idea from Coqui there should be a way to identify deep-fakes in voice context. After some Twitter chatting[1] there seems one thing without doubt: "It's the old story between hacker and the people trying to prevent misusage".
Possible techniques[2][edit | edit source]
What kind of techniques are useful for what and what's the pros and cons:
Watermark in TTS output[edit | edit source]
Easy to analyse / reproduce using original sourcecode.
Watermark in TTS dataset[edit | edit source]
Models can learn to reproduce watermark without seeing anything on that in the code.