AI Transcription 2.16 — Mac version

Journalists, students, and anyone who hates typing up interviews end up here. AI Transcription turns audio and video files into text without sending anything to the cloud. Audio never leaves the Mac.

Drag a voice memo in to see text appear

• Transcribe a short audio file. Drag in a voice memo or a clip from a meeting and watch the text appear. The app runs Apple Speech Recognition and OpenAI Whisper directly on your Mac.
• Split the transcript. You can break the text by words, sentences, or duration. Useful when you want readable paragraphs instead of one long wall of text.
• Export as SRT or WEBVTT. If the file is for subtitles, these formats drop straight into most video editors. Plain text and CSV are there too.

Why it suits a first run

There's no account to set up and no upload wait. You pick a file, the Mac does the work, and the transcript lands in a format you can actually use. Version 2.16 runs on macOS 14.6 and supports both Intel and Apple Silicon Macs.

Engine switching feels snappier now

Version 2.16 tidied up the transcription engine under the hood. The app still uses Whisper and Apple Speech Recognition, but the switch between them feels snappier now. You pick a model, drop a file, and it gets on with it.

Earlier builds like v2.12 and v2.13 introduced better handling of long recordings. Before those updates, a two-hour interview might stutter on an older Intel machine. The ai transcription 2.12 release tightened memory use, and 2.13 carried that forward. By 2.15, the export panel gained small quality-of-life touches — SRT timestamps got more accurate around speaker changes, and CSV exports no longer mangle punctuation.

Test the same clip against two engines

• Try both engines on the same short clip. Whisper sometimes catches words Apple Speech Recognition misses, and the other way round. A one-minute voice memo is enough to compare. The app switches engines in a couple of clicks.
• Name your files before you drag them in. The app uses the filename in the transcript header. A clear name now saves renaming exports later.
• Check the split setting early. The default split by sentences suits most jobs, but if you're exporting subtitles switch to duration-based splitting before you export. It saves rework.
• Keep the app open while it works. It's not a background service. If you close the window mid-transcription, you lose progress.

If it won't open

macOS Gatekeeper blocks apps from outside the App Store. Mount the disk image, then open Terminal and type xattr -cr followed by a space. Drag the app icon into the Terminal window, press Return, and launch it from Applications as normal.

Transcribing a team meeting: start to finish

Say you recorded a forty-minute team catch-up on your phone as an M4A file. You need action points, not a word-for-word script.

1. Get the file onto the Mac. AirDrop it or pull it from iCloud. Drop it straight into the AI Transcription window.
2. Pick the engine. For multi-speaker recordings, Whisper tends to handle overlapping voices better. Select it from the dropdown and hit start.
3. Read through the raw transcript. Scroll down and clean up any speaker names or mumbled bits. The text is editable right in the window.
4. Split by sentences. That gives you readable paragraphs instead of one block. Switch the split mode in the sidebar.
5. Export as plain text. Copy the text into your notes app or team wiki. If someone wants subtitles for the recording later, export again as WEBVTT — the original file stays untouched, so you can export as many formats as you need.

The whole job, from AirDrop to finished notes, takes about as long as the recording itself on an Apple Silicon Mac. On an older Intel machine, allow a little more time and close other apps that are chewing through memory.

• macOS 14.6 or later
• Apple Silicon or Intel 64-bit Mac