How do you transcribe a QuickTime audio or screen recording?

To transcribe a QuickTime recording, get its audio into a format your tool accepts, run the tool, then read the draft against the recording. An audio-only recording usually needs no change. A screen recording is a video file, so for most cloud tools you pull out the audio track first. Whatever the route, listen to the places where the draft looks wrong, because a machine draft is a starting point, not a finished transcript.

An audio recording is usually ready to transcribe as it is

QuickTime Player records audio in two quality settings. Apple says the High setting records 44100 Hz AAC audio, and the Maximum setting records uncompressed 24-bit AIFF-C audio ¹. Apple's steps finish with File > Save, and the page does not name the file extension, so read the name in the Save dialog.

An .m4a file normally holds AAC audio. OpenAI's transcription API accepts m4a, along with mp3, mp4, mpeg, mpga, wav and webm ². Microsoft's Transcribe feature in Word takes .wav, .mp4, .m4a and .mp3 files ³. If your file has one of these extensions, send it to the tool as it is.

Maximum-quality recordings are a different case. Apple warns that they "produce uncompressed files, which can use large amounts of storage space" ⁴. A long recording at that setting can be too large for a cloud upload, so check the size before you send it.

A screen recording is a movie, so export its audio first

Apple says a screen recording is saved as "an H.264 format QuickTime movie with an AAC audio track (if audio was selected)" ⁵. The audio is inside a video file, and the video takes up most of the size. Neither OpenAI's list nor Word's list names a .mov file ^{2 3}, so a screen recording usually has to be converted before a cloud tool will take it.

The simplest fix is to keep only the audio. With the free ffmpeg command-line tool installed, this command writes the audio track to an .m4a file without re-encoding it:

ffmpeg -i recording.mov -vn -c:a copy audio.m4a

The -vn option drops the video, and -c:a copy keeps the AAC audio as it is. If the copy fails, re-encode the audio with -c:a aac in place of -c:a copy. Keep the original .mov so you can check the transcript against it.

Free tools on your own computer can read the movie directly

Open-source Whisper is one route that needs no upload. Whisper's audio loader starts the ffmpeg command-line tool to decode each file, and it needs ffmpeg in the PATH ⁶. Because ffmpeg reads the movie's audio, Whisper should accept a .mov file without an export step. We have not tested a QuickTime file in this way, so run a short clip first.

Whisper runs on your own computer, so the recording does not leave it. That matters for a private recording. It also takes setup, and a slow computer may need a long time for an hour of audio.

Cloud tools cost a few cents an hour but have a size limit

Cloud transcription is cheaper than many people expect. At OpenAI's listed price, one hour of audio with gpt-4o-mini-transcribe costs about 0.18 US dollars ⁷. Price is not the only limit. OpenAI's Transcriptions API accepts files up to 25 MB, and a larger recording must be compressed or split into chunks ⁸.

A screen recording is often over that limit, because of its video. Export the audio first, which shrinks it a lot, and then check the size. If the audio is still too big, split it at pauses between sentences, and check the words at each cut against the sound.

It depends on the recording and what you need the transcript for

• A private recording should stay on your own computer, so use Whisper or another offline tool. The question on offline transcription covers that route.
• A free transcript on a short clip is easy with a cloud service's free tier, but read its terms about how it stores audio. The question on free transcription compares the options.
• A noisy or faint recording produces more errors, however you export it. The question on low-quality audio explains what to do before you transcribe.
• More than one speaker needs speaker labels, which many tools do not add by default. See how to transcribe audio with multiple speakers.

Steps to transcribe a QuickTime recording

1. Check the file type. Audio-only recordings are usually AAC in an .m4a file. A screen recording is a .mov movie.
2. Export the audio if needed. Use the ffmpeg command above for a screen recording, or keep the .m4a file as it is.
3. Test a short clip. Transcribe ten seconds of clear speech, and compare the text with the sound.
4. Transcribe the whole file. Split it at pauses if it is over the tool's size limit.
5. Proofread against the recording. Check names, numbers and any quiet passages, since those are where errors gather. The guide to transcribing audio to text covers the proofreading steps in more detail, and the format question lists more formats.

The overview of transcription covers the other routes, including video and meeting recordings. Accuracy depends on the tool and the audio. Vendors and benchmarks report it as a word error rate, the share of words a tool gets wrong, but that number comes from clear speech, so your own check of the transcript still matters.