Transcribe MP4

MP4 is 15.9% of everything uploaded here — 5,987 files, median 13 minutes 49 seconds at 684 kbps. Almost all of that bitrate is picture, and none of it is transcribed: the audio track is extracted on arrival and the video is discarded.

Two things follow. Resolution and codec make no difference to the transcript — a 4K H.265 clip and a 720p H.264 clip of the same length take the same time and produce the same text. And if upload speed is your bottleneck, exporting the audio alone cuts the file by roughly 90% without touching a single sample of sound.

These are medians measured on our production cluster, expressed as time per hour of audio.

Against the MP4 files uploaded here: the median, just under 14 minutes, comes back in about 20 seconds. Nine in ten are under 67 minutes, roughly a minute and a half. The video track adds a little while the audio is extracted, and nothing after that.

Not for transcription. MP4 was standardised from the QuickTime file format, so the two containers share the same structure and both upload here as they are, along with .m4v and .m4a.

The practical difference is what tends to be inside. MOV often carries ProRes and uncompressed audio, which is why MOV files hit the size limit sooner; MP4 is usually H.264 or H.265 with AAC, which is why the median MP4 here is 47 MB against MOV’s 54 MB despite similar durations.

For MP4 it is usually size. At the 684 kbps median we see, 5 GB is around 16 hours — but that median hides a wide spread, and the largest MP4 processed here was 4.78 GB, just under the ceiling. A high-bitrate 4K recording can reach 5 GB in under an hour.

• Export the audio if the file is close to the limit. An hour of MP4 at our median bitrate is about 290 MB; the same hour as audio alone is around 60 MB.
• Split at scene boundaries, not by the clock. Cutting mid-sentence costs the sentence.
• Keep the original. A transcript can be re-run; a re-encoded master cannot be restored.