Wouldn’t it? Something like Whisper is great for recognition, and is built on a transformer architecture, like most of the SOTA voice stuff is.