Easier to wait for, easier to read. • See how long a transcription will take, or choose Do this later. • A transcription that didn't finish can be retried for free. • Screens fit large text and landscape; iPad shows list and transcript side by side. • Players show your recording's real waveform. • Errors, the widget and Siri now speak your language.

Sharper transcripts, in every language. • Long recordings no longer repeat a section of the transcript. • Sentences now end with the right mark for the language you speak, instead of a full stop everywhere. • Removing filler words no longer deletes the space French and Greek put before ? and !. • A line that already ends in a quotation mark no longer gains a second full stop.

Faster and lighter, especially on long recordings. • Summaries of long recordings finish sooner. • Copying a long transcript no longer freezes the screen. • Opening a recording to play it is smoother. • Long recordings use less memory while transcribing. • Less work at startup, so the app stays responsive.

Recordings you don't lose, and imports that finish in the background. • A recording cut short — by a device running out of space, say — now keeps everything captured up to that point. • Imported audio keeps transcribing after you switch apps. • Imports are faster, and space from interrupted ones is reclaimed. • Speaker labels is marked Pro in Settings, so you know before recording.

Faster transcription, and it finishes even when you leave. • New: skip silence — long recordings with quiet stretches transcribe much sooner. Switch it off in Settings if you speak softly. • Everything transcribes faster, markedly so on older phones. • Queue several recordings and switch apps: they all finish now, not just the first. • A stalled transcription can be cancelled.

Video import, and live text while you record. • Import a video and transcribe its audio — no need to pull the sound out first. • Watch words appear as you speak. Like everything else, it stays on your device. • Start recording right away while a speech model is still downloading. • A new app icon and launch screen, in light and dark.

Faster, steadier transcription. • Speaker identification is much faster — the recognition pass now runs across several CPU cores instead of just one. • iPads with more CPU cores now put them to work during transcription. iPhone performance is unchanged. • Deleting a recording while it is still transcribing now stops that work immediately instead of letting it run on in the background. • Fixed a rare case where an interrupted transcription could be saved as an empty transcript, or quietly start over the next time you opened the app.