Use case
Multilingual voice dictation
OSSFlow does not ship a list of 100 languages, and it does not translate. The default Grok setup sends language en. Another OpenAI-compatible model can transcribe other languages when that model supports them. You still read the paste.
What the default does
The Grok guide leaves Language as en and Model as grok-voice-transcribe-2.0. That is English transcription, not a translator. Custom cleanup can ask for a dialect of English, such as British English. A dialect instruction does not turn French speech into English text.
The personal dictionary stores spellings. It does not store translations.
When another host is the right tool
Groq's speech-to-text docs describe whisper-large-v3 and whisper-large-v3-turbo as multilingual. OSSFlow can call them if you set the OpenAI-compatible base URL to https://api.groq.com/openai/v1 and set the model id. The API key guide has the fields. OSSFlow still does not call Groq's translation endpoint, so a French recording is not rewritten into English by OSSFlow.
Gboard, Samsung voice input, SwiftKey, and Voice Access each publish their own language notes. Those pages are linked from the Android roundup. Wispr Flow's Android page says 100+ languages. That claim is theirs. It is not an OSSFlow feature.
Questions
- Does OSSFlow auto-detect the language?
- Not as a feature of the app. The Grok default sends en. A provider model may detect language on its own. OSSFlow does not promise that, and it does not translate.
- Can I dictate in English and get another language out?
- No. OSSFlow pastes a transcript. It does not translate.
- How do I ask for British English?
- Leave the Grok language as en, and put British English in Custom Cleanup instructions. That asks the cleanup model for a dialect. It is still English.
- Does OSSFlow support 100 languages?
- No. That figure is on Wispr Flow's Android page, and this site does not copy it onto OSSFlow. Check the model provider for the languages that model documents.