v1.5.6
What you said belongs to you, and only you.
Privacy: cloud recordings are deleted immediately once transcription finishes. Async transcription models need audio uploaded to the cloud for processing first. Starting with this release, the moment transcription finishes, Voilà immediately deletes the cloud recording file and the transcription job record — no copy of your voice stays on any server.
Apple's on-device live preview feels more responsive. Preview text used to animate in typewriter-style, which felt laggy on long sentences; it now shows recognized text the instant it arrives, keeping pace with your speech. Also fixed a potential crash when the system speech recognizer doesn't support the selected language — you now get a clear message instead.
Soniox fully switched to v5. The server has retired the v4 model, so this release removes the v4 option from the client entirely and fully aligns with the v5 interface. Any leftover v4 setting in your existing config gets migrated to v5 automatically on launch, no manual steps needed.
"Voilà" gets recognized correctly now. Common mis-hearings of the brand name (Viola, Vola, Wala...) are now automatically corrected to Voilà.
No settings changes needed — just update and go.