⏳ A next-generation speech generator has emerged — Dessa. Jack Dorsey’s company Block, founded by one of the creators of Twitter (X), has developed the world's first model trained exclusively on real human conversations by listening, without using any text. It is currently the AI model with the most realistic speech. Dessa can listen and speak simultaneously, laugh in the middle of a sentence, gasp in surprise, and carry on conversations while interpreting vocal intonation. It can even recognize sarcasm. The model can run locally on a laptop like a MacBook Pro. However, it is currently available only in a limited preview for researchers. In the first video, Dessa has a conversation with ChatGPT — and the difference in how they sound is immediately noticeable.
- 3
- 2
- 1
- 1