World's First Dessa Speech Generator Trained on Live Dialogues
Dessa, a speech generator trained on live dialogues without text labels, is now available.
Model Features
Unlike competitors, Dessa can interrupt its conversation partner and responds instantly to their phrases. The model recognizes sarcasm and uses natural interjections to maintain the flow of conversation.
Application and Access Restrictions
- The model was created by Twitter's co-founder
- Open access to the technology is currently restricted
- Developers fear misuse by fraudsters
Experts consider the technology too realistic for public distribution in its current form due to the risks of creating fake voice dialogues.








