What LUFS Should Video Be? Loudness, True Peak and Timing Basics for Video

What LUFS Should Video Be? Loudness, True Peak and Timing Basics for Video

For online video, a widely used target is about -14 LUFS integrated loudness with a true peak no higher than -1 dBTP. Broadcast is quieter: EBU R128 recommends -23 LUFS. Measure the whole finished mix, not individual clips, and line up audio cues to whole video frames using the frame rate and tempo.

The rest of this post explains what those numbers mean, how to measure them for free, and how to convert between frames, beats and samples.

What is LUFS?

LUFS (Loudness Units relative to Full Scale) is a unit of perceived loudness, defined by the measurement method in ITU-R BS.1770, that weights frequencies roughly the way human hearing does and ignores near-silent passages.

LKFS is the same unit under a different name, used mainly in North American broadcast specs.

Integrated loudness is the average loudness of an entire file or program, measured in LUFS with gating so silence doesn't pull the number down.

BS.1770 meters also report momentary loudness (a 400 ms window) and short-term loudness (a 3 s window), which help you spot loud moments inside a mix.

Why is -14 LUFS the common target for online video?

Many streaming and video platforms turn down audio that is much louder than their reference, so mixing far above it mostly costs you dynamics. Around -14 LUFS integrated has become the common reference for online video and streaming.

Treat it as a starting point. Platforms don't all publish one fixed number. YouTube, for example, says it may apply audio enhancements such as automatic volume adjustments and Stable volume. Some editors mix short-form vertical video slightly quieter, around -16 LUFS, for phone speakers. What matters most is a clear voice, no clipping, and consistency across your uploads.

What do EBU R128 and true peak mean?

EBU R128 is the European Broadcasting Union's loudness recommendation. It sets a program target of -23 LUFS and a maximum true peak level of -1 dBTP, and adds a Loudness Range (LRA) descriptor. The full text is in the R128 PDF.

True peak is the estimated maximum level of the reconstructed analog waveform, including peaks that fall between digital samples, measured in dBTP (decibels relative to full scale, true peak).

Sample peaks can read -0.3 dBFS while the true peak is above 0 dBTP. Lossy encoders (AAC, MP3, Opus) can raise peaks further, which is why -1 dBTP is a common ceiling: it leaves room for the encode.

Loudness Range (LRA) describes how much the loudness varies over a program, in LU. Dialogue-heavy content usually has a lower LRA than a film score.

How do I measure LUFS and true peak with ffmpeg?

ffmpeg's ebur128 filter measures integrated loudness, LRA and true peak:

ffmpeg -hide_banner -nostats -i final_mix.wav -af ebur128=peak=true -f null -

Read the summary at the end: I: is integrated loudness in LUFS and the True peak block shows the peak in dBFS (true-peak).

To normalize toward a target, the loudnorm filter implements EBU R128-style normalization. The usual approach is two passes: measure first, then apply with the measured values.

# pass 1: measure
ffmpeg -i in.wav -af loudnorm=I=-14:TP=-1:LRA=11:print_format=json -f null -
# pass 2: apply, pasting the measured_* values from pass 1
ffmpeg -i in.wav -af loudnorm=I=-14:TP=-1:LRA=11:measured_I=-18.2:measured_TP=-3.1:measured_LRA=6.4:measured_thresh=-28.5:linear=true -ar 48000 out.wav

The measured_* numbers above are placeholders. Replace them with your own pass-1 output. In Python, pyloudnorm gives the same kind of BS.1770 measurement.

How do I convert frames, beats and samples?

Three formulas cover almost every timing job:

• Seconds per frame = 1 / fps
• Samples per frame = sample rate / fps
• Frames per beat = fps × 60 / BPM

At 48 kHz, samples per frame are:

Frame rate	Samples per frame
24 fps	2,000
25 fps	1,920
30 fps	1,600
60 fps	800
29.97 fps	1,601.6 (not a whole number)

Frames per beat at 30 fps:

BPM	Frames per beat
90	20
100	18
120	15
80	22.5
128	14.06

Tempos that give whole frames per beat (90, 100, 120 at 30 fps; 120 at 24 fps gives 12) make it easy to cut picture on the beat. Tempos like 80 or 128 BPM at 30 fps drift against the frame grid, so put hard cues on the nearest whole frame and accept a few milliseconds of offset.

A cue frame is the specific video frame number where an audio event (a hit, a drop, a logo sting) must land.

A countdown makes cue frames easy to hear: in this one, each pip lands exactly on a whole second (frames 0, 30, 60 and so on at 30 fps).

How do I put these numbers into a sound prompt?

If you generate audio from code, write the numbers straight into the prompt. Here is a drop that lands on a fixed frame, the same structure as the Heavy impact drop entry on Opus Sounds Directory:

Make a 3 s transition drop for video at 30 fps.
Output: one WAV, 48 kHz / 16-bit / stereo, exactly 144000 samples.
Python + numpy/scipy only. Synthesise everything (no samples, no downloads). Fixed random seed.
Sound: sub drop from 120 Hz to 40 Hz + short noise burst + 1.2 s tail.
Peak impact aligned to frame 45 (1.5 s, sample 72000).
Master: -14 LUFS integrated, true peak <= -1 dBTP.
Verify length, loudness and peak with ffmpeg ebur128; confirm the RMS maximum falls in frame 45;
render a spectrogram. Fix and re-render until every check passes, then report the numbers.

And a 30-second bed on a 90 BPM grid, which divides evenly into 20 frames per beat (compare the Traffic glitch → warm strings entry):

Make a 30 s music bed for a vertical video ad at 30 fps.
Output: one WAV, 48 kHz / 16-bit / stereo, exactly 1440000 samples.
Python + numpy/scipy only. Synthesise everything. Fixed random seed.
Timing: 90 BPM (1 beat = 20 frames, 1 bar = 80 frames). Cues: frame 0 = hit;
frame 320 (bar 5) = texture change; frame 800 (bar 11) = full arrangement;
last bar ends with a soft final chord that fades to exactly 0 at sample 1440000.
Sound: slow legato strings, soft sub bass, gentle pulse, key A minor.
Master: -14 LUFS integrated, true peak <= -1 dBTP.
Verify length, loudness and peak with ffmpeg ebur128 and report the numbers.

Opus Sounds Directory shows each entry's target next to its measured integrated LUFS and true peak, which is a handy way to see how far a render can land from the number in its prompt. Always measure the final mix yourself.

What's a simple loudness checklist before export?

1. Mix with voice first, then bring music and effects in underneath.
2. Measure the full mix with ebur128.
3. Adjust gain toward your target (about -14 LUFS for online, -23 LUFS for EBU broadcast).
4. Keep true peak at or below -1 dBTP.
5. Check that cues land on their frames.
6. Listen on a phone and on headphones.