Hey everyone,
I’ve noticed a regression in Codex voice input lately and wanted to see if others are experiencing the same thing.
The main issue right now is unclear mic feedback, where it is really difficult to tell whether the microphone is actually capturing audio correctly while speaking. On top of that, transcription takes noticeably longer after stopping the recording compared to ChatGPT, and the resulting text has become less precise with more errors overall.
Previously, things like the active small loading indicator, overall transcription speed, and recognition accuracy were all much smoother.
Curious if anyone else has run into this.