# Diktator 2.0.0 — Build 45

Build 45 improves segmented dictation when a speaker switches languages during
one recording.

- Detects the language of each speech segment independently when more than one
  dictation language is selected.
- Avoids carrying a text prompt from one language into the following segment,
  reducing omissions and substitutions at language boundaries.
- Applies language-specific post-processing to each detected segment.
- Allows shorter two-second segments for brief words and phrases between
  pauses, while preserving the existing default for current users.
- Keeps single-language dictation behavior unchanged.

Speech recognition remains on the Mac. Network access is used only when the
user downloads a model, requests an update check, or uses licensing and
purchase features.