TTS watermarks: Difference between revisions
Jump to navigation
Jump to search
m (Added category) |
m (Added new thoughts from Erogol about watermarks.) |
||
| (One intermediate revision by the same user not shown) | |||
| Line 2: | Line 2: | ||
== Preventing deep fakes == | == Preventing deep fakes == | ||
Based on Erogol's idea from Coqui. | Based on Erogol's idea from Coqui there should be a way to identify deep-fakes in voice context. After some Twitter chatting<ref>https://twitter.com/erogol/status/1464412634783043585?s=20</ref> there seems one thing without doubt: "It's the old story between hacker and the people trying to prevent misusage". | ||
==Possible techniques<ref>https://github.com/coqui-ai/TTS/discussions/1036#discussioncomment-1863431</ref>== | |||
What kind of techniques are useful for what and what's the pros and cons: | |||
=== Watermark in TTS output === | |||
Easy to analyse / reproduce using original sourcecode. | |||
=== Watermark in TTS dataset === | |||
Models can learn to reproduce watermark without seeing anything on that in the code. | |||
== References == | |||
<references /> | |||
Latest revision as of 18:49, 23 December 2021
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.