How to apply this guide
Separate local speech recognition from optional writing AI and recover the original when formatting is unavailable.
Start with
Choose a working local speech model and record a short sentence. Read the original. It should be possible to complete plain dictation without an optional writing provider if other access requirements are met.
Check before you finish
Use local Dictation when a complete transcript matters most. Try rewriting after the provider reconnects, and compare the edit against the saved original.
Evidence to keep: Keep the target field open and verify that the complete reviewed result reached the intended destination. This page’s key limit is: Separate local speech recognition from optional writing AI and recover the original when formatting is unavailable.
Start here
Whisper or another local speech engine creates text from audio. Ollama can run a separate writing model. Stopping one service does not automatically stop the other stage.
Test plain Dictation
Choose a working local speech model and record a short sentence. Read the original. It should be possible to complete plain dictation without an optional writing provider if other access requirements are met.
Test writing separately
Switch to a mode that uses your configured provider. If it fails, check that Ollama is running and that the selected model exists. For Email, check the local layout or copy the original if a rewrite failed.
Leave a reliable fallback
Use local Dictation when a complete transcript matters most. Try rewriting after the provider reconnects, and compare the edit against the saved original.




