In the resulting transcript, most timestamps are OK, but a few timestamps are regularly wrong. These incorrect timestamps have in common: People in the clips say Aaah or Hmmm too long before speaking (one speaker, for 4.5 seconds). Do others in the community also have the same problem? If “Yes,” I suggest that the speech-recognizing AI also recognizes too long Aaah or Hmmm in the future (or, the possibility of manually correcting the few incorrect timestamps).
AdChoices