Soniox moves to STT v6 and gets the features of the macOS plugin (#735). Soniox retires v5 on December 15, 2026 and routes v5 requests to v6 from then on.
- Model: Automatic follows Soniox's newest models, live text and recordings each on their own. The plugin loads the model list for your region, keeps it for offline use and refreshes it once a day; Refresh models loads it on demand. Choose a version such as v5 to stay on it.
- Transcription context: an optional text with topic, names or situation goes to Soniox with every transcription, together with your dictionary terms.
- Translation to English for recordings and files, with subtitle timing. English passages in mixed speech stay as spoken.
- Speech output: Soniox can now read text aloud for spoken feedback, with its own voices (default Adrian) and a choice of speech model.
- Recordings longer than 270 minutes are split before upload.
- The region setting moved to the connection settings, and the settings are in English and German.
v6 writes amounts closer to speech: "1.234 Euro 56" where v5 wrote "1.234,56 €". Choose v5 to keep the old format until Soniox retires it.
Requires plugin host 1.1.5, which every TypeWhisper for Windows from 1.1.0 provides.
Source commit: d4e8e8f