# Diktator 2.1 — Build 61 Diktator 2.1 expands local multilingual work while keeping speech recognition on the Mac. - Choose a translation target from the full language catalog. Translation with the selected local speech model is experimental: quality varies by model and language pair. - Optionally translate completed text with Apple Translation. Apple language downloads and On-Device Mode are managed in macOS Translation settings. - Use a portable post-processing pack with 13 profiled languages, now including Arabic and Turkish. - Improve Chinese and Japanese plausibility checks and apply language-specific processing independently to mixed-language speech segments. - Use the complete language catalog for dictation instead of being limited to the 13 profiled interface languages. - Improve the translation and language menus, Accessibility-aware delivery, onboarding, reference windows, and the nonblocking About panel. - Add the verified Russian model mirror as the final fallback after the existing international model sources. Local speech-model translation is a practical experimental feature, not a guarantee that every model will translate every language pair equally well. Network access is used only for model and update downloads, licensing and purchase features, and Apple Translation when macOS On-Device Mode is not used.