Delivery QC

Delivery QC

mj qc render.mov                    against your qc_spec setting, else "web"
mj qc render.mov broadcast-us       a built-in spec
mj qc render.mov client.mjspec      a spec file
mj media.qc path=/abs/render.mov format=broadcast-eu

Exits 1 when a check fails, so mj batch and scripts can stop a bad file going out. Stock macOS only: avmediainfo reads the movie, afconvert decodes the audio.

Built-in specs

Spec	What it checks
broadcast-us	mov, ProRes, 1920x1080, 29.97 or 23.976 fps, stereo 48 kHz audio, -24 LUFS +/- 2, peak at most -2 dBFS, colour tags
broadcast-eu	mov, ProRes, 1920x1080, 25 fps, stereo 48 kHz audio, -23 LUFS +/- 1, peak at most -1 dBFS, colour tags
web	mp4 or mov, H.264 or HEVC, 48 or 44.1 kHz, -14 LUFS +/- 2, peak at most -1 dBFS
social-vertical	mp4, H.264, 1080x1920, 23.976-30 fps, at most 90 s, -14 LUFS +/- 2
prores-master	mov, ProRes, 48 kHz audio if any, colour tags

These are starting points; check your client's sheet.

Spec files

key = value lines, # comments on their own lines (a comment after a value becomes part of the value). Lists are comma-separated.

name = Client X master
container = mov
codec = prores
width = 3840
height = 2160
fps = 23.976, 24
minDuration = 15
maxDuration = 30.5
audio = required
audioChannels = 2
audioSampleRate = 48000
loudness = -24
loudnessTolerance = 1
peakMax = -2
colorTags = required

Key	Values
container	file extensions: mov, mp4, m4v, ...
codec	prores, h264, hevc, mjpeg, png, animation, mpeg4
width, height	pixels
fps	one or more frame rates
minDuration, maxDuration	seconds
audio	required, none or any
audioChannels, audioSampleRate	a number (a list for the sample rate)
loudness, loudnessTolerance	integrated LUFS, and how far off is allowed (default 1; 0 means exact)
peakMax	dBFS
colorTags	required or any

Every value is checked before the movie is read. A typo such as audio = requried, a word where a number belongs, or a spec that asks for nothing is an INVALID_SPEC error, never a silent pass.

How loudness is measured

ITU-R BS.1770-4 K-weighting with EBU R128 gating (absolute -70 LUFS, relative -10 LU), computed in Python on the decoded audio. It agrees with ffmpeg's ebur128 meter within 0.1 LU in the test suite. The peak is the sample peak; a true-peak meter can read up to about 0.5 dB higher on bright material, so leave headroom. Black and frozen frames are not checked. Audio shorter than 0.4 s cannot be measured and is reported as not measured. The audio is decoded to a temporary folder first (about 11 MB per minute of stereo 48 kHz); if that does not fit in free space the loudness checks are skipped and say why.