Guides · Offline Transcription

Record the call with the iPhone's built-in recorder, then decide whether Apple's transcript is enough or whether you need your own.

The free, built-in, open-source and paid ways to transcribe on your own device, and which one fits which job.

What determines local transcription speed, and what to change when it is too slow.

The Neural Engine, the GPU, quantised models and voice detection — what each part actually does.

The cause of Whisper's repeated-phrase loop, and four ways to get rid of it.

What each transcript and subtitle format is for, and which one your task actually needs.

Getting text out of a meeting recording, including the recordings the built-in transcription will not touch.

Three ways to turn an iPhone recording into text, including the one that works in airplane mode.

What speaker diarisation can and cannot do, and the workflow that does not depend on it.

What actually goes wrong with long recordings, and how to get a clean transcript of one.

Turning recorded lectures into searchable, quotable notes — including the part where you ask permission.

A working method for journalists, researchers and anyone who has to stand behind a quote.

Producing episode transcripts, and turning them into show notes, clips and search traffic.

The things that genuinely improve transcription accuracy, in order of how much they matter.

What each common audio and video format means for transcription, and the ones that genuinely cause trouble.

Three ways to prove for yourself whether a transcription app sends your recordings anywhere.

Turn a recording into a timed SRT file, then load it into Final Cut Pro, Premiere or YouTube.

The honest trade-off between Whisper model sizes, and when the small ones are enough.