CommandSmith 0.3.78
Voice transcription
- Dictation starts recording right away. Switching to a profile's model, warming the model up and reading the screen context now happen while you speak, not before recording starts.
- The two most recently used models stay loaded, so switching between two profiles' models is instant.
- Pressing Esc discards a dictation immediately, so dictating again right after no longer waits for the discarded recording to be transcribed.
- The first dictation after a break is faster: the engine is no longer slowed down by App Nap while idle, and the model warms up as you start speaking.