Choose the transcription model that fits your language and workflow, and download only what you need. We’ve also improved audio handling to make dictation more reliable.