Arabic voice typing
Use the shared All languages model.
Speech recognition languages
Saydraft supports multilingual dictation on Windows through local speech models. Thai, Chinese, Korean, Japanese, Hindi, Urdu, Indonesian, Malay, Spanish, Brazilian Portuguese, Bengali and Arabic are among the available language choices. Models have different coverage; choose the one that fits the languages you speak.
Use the shared All languages model.
Use the shared All languages model.
Brazilian Portuguese is covered by the default model and All languages.
Use the shared All languages model.
Use the shared All languages model.
French is covered by the default model and All languages.
German is covered by the default model and All languages.
Use the shared All languages model.
Use the shared All languages model.
Use the shared All languages model.
Use the shared All languages model.
Use the shared All languages model.
Use the shared All languages model.
Spanish is covered by the default model and All languages.
Use the shared All languages model.
Use the shared All languages model.
Use the shared All languages model.
Use the shared All languages model.
Use the shared All languages model.
The optional All languages model is Whisper large-v3-turbo. A single approximately 1,621.7 MB installation covers Thai, Mandarin Chinese, Korean, Bengali, Arabic and other supported Whisper languages, including Japanese, Hindi, Urdu, Vietnamese, Filipino / Tagalog, Tamil, Telugu, Persian, Indonesian and Malay. There are no separate language packs.
The default Parakeet v3 model covers 25 European languages, including Spanish, French, German and Portuguese. Brazilian Portuguese uses the Portuguese choice, code pt; there is no separate Brazil-only decoder setting.
Selecting a language starts no download. If you keep manual model selection enabled, choose a compatible model yourself. For direct languages outside Parakeet’s set, pin the language with automatic routing or manually choose All languages.
These are speech recognition choices: speak a language and receive its transcript. They are not translated app menus or additional Translate-mode output targets. Recognition quality varies with accent, audio and model; language coverage is not an accuracy guarantee. Review your text, especially names and mixed-language speech.
Get Saydraft and review setup · Follow the first-use guide · Read the language and model reference ↗