Skip to main content

Language Support

Phlox can transcribe audio and generate clinical notes, letters, and chat responses in your working language. Each user picks their own preferred language in Settings → User Settings → General, and it drives transcription, LLM output, and date formatting for their account — clinicians in the same practice can each work in a different language.

What the language setting controls​

  • Transcription — the language hint passed to the speech-to-text engine.
  • Output — notes, letters, summaries, and chat are produced in the selected language.
  • Formatting — dates render in the matching locale.

Transcription models​

The transcription language you can actually use depends on your speech-to-text engine:

  • Desktop (bundled engine) — the default Omi Med STT v1 model is English-only. When you select a non-English language, Phlox offers to download the multilingual Parakeet TDT 0.6B v3 model, which understands 25 European languages and auto-detects the spoken language. Downloading it switches it to the active model. Until it is active, non-English recordings fall back to English.
  • Remote (Whisper-compatible) — point Phlox at any endpoint in Settings → Admin Settings → Whisper; remote endpoints generally support any language.

Live agent sessions use the same engine as recordings. On Docker, setting WHISPER_STREAM_URL routes live utterances to a separate endpoint (see Setup) — that endpoint receives the language hint the same way.

See Transcription for the full recording and dictation workflow.

Notes, letters, and chat​

Output language is applied automatically — no per-template or per-prompt configuration is required. The model is instructed to write every output in your selected language, regardless of the language of the source audio or documents.

Application interface​

The Phlox interface is currently English-only — the language selector lists only languages with a complete interface translation, so English is the only option today. Multilingual transcription and output are fully implemented and activate as soon as your language appears in the selector. Interface translations are contributed by the community; if you'd like to add yours, see Contributing → Adding a language.