Documentation / Help
Slow dictation
Text should appear about a second after you let go. If it takes much longer, one of these is usually the cause.
The speech model
A bigger model is more accurate and slower, and an older computer feels it most. Under Speech model in Settings, choose Fast (Whisper base) to speed things up, or Balanced (the default) if you are on Accurate. On a Mac, speech recognition uses the graphics chip; on Windows it uses the processor.
The AI model
The clean-up adds the time your provider needs to answer. A small, fast model keeps it under a second. A large one, or a model on a slow computer, takes longer. If the provider does not answer in time, SayFeather inserts the text as heard and says AI too slow. Pick a smaller model under AI in Settings. To see the difference, Try it shows the time in milliseconds.
The language
If you always speak one language, choose it under Language you speak instead of Detect automatically. That is a little faster.
On Android
Phones are slower than computers. Stay with Fast unless you need the accuracy; Accurate can take many seconds for a short sentence.
Stuck? Email [email protected].