Using beta 1.11.0b1 - and “Extract Pitch From Audio” on isolated vocals from some commercial tracks, assigned to Synth V singers, here are some examples of what you get:
My first impression is that the pitch extraction is just TOO good, almost exactly like the original recording. But the lyric recognition is not very good. I had to re-type most of the words (but, of course, that doesn’t take too much effort.) I tested the same files in Ace Studio, and the lyric detection was better, and I only had to correct about 20% of them. But the pitch extraction in Synth V is much better than in Ace Studio. You can fix lyrics, but it takes a lot of time to fix pitch extraction, if not impossible, so this is great. Dreamtonics will surely improve lyric detection in the future.
Hearing these examples you may think the tracks are the original singer, but here are the raw vocal files from Synth V, so you can hear that they are in fact the AI singers: