Transcription

Apple speech recognition turned into text — on-device first, optional Apple-server jobs, your edits on top, and TXT/SRT export.

Transcription

Transcription turns speech in a recording into written text using Apple's speech recognition. You start it yourself for one recording — there is no automatic mode.

Language: version 1 writes out en-US English. You can record and export audio in any language, but there is no automatic language detection, translation, summary, or "who spoke when" labeling.
On your phone first: the app always tries on-device recognition first (SpeechAnalyzer on supported iOS 26 and later, Apple's on-device recognition on iOS 17–18). On-device work never leaves your phone.
Apple's servers as a fallback: if on-device recognition isn't available for your phone and setup, you can choose to send that one recording to Apple's speech servers. This needs a setting you turn on plus a confirmation each time — there is never a silent switch, and the app tells you when a job would use the server.
Your edits win: Apple's text is the starting point. Your corrections are kept on top and never quietly overwritten.
Views and export: read the transcript as plain paragraphs (default) or with timestamps. TXT export is the plain visible text; SRT keeps the timestamps.

Transcript — plain paragraphs or timestamped, editable, exportable

Long recordings are handled in short pieces so a long file doesn't fail or run out of memory.
Honest job status: each job shows where it ran (on-device or server), its progress, and whether it was cancelled, failed, or can be retried. Nothing shows fake progress while the system has paused it.