Re:Seamless: Meta's New Speech Models
by _kb on Monday December 04, 2023 @04:45AM
(#38515405)
Apart from the dynamic range challenges for sensing, source separation is hard. There's been a pretty long line of research into the area - see "cocktail-party problem". AFAIK it's still a mostly unsolved problem.
[ Parent
| on HN ]
There are no comments on this comment yet.