How Loudness Works: LUFS, dBFS and Why Your Mix Sounds Quiet

Many new editors judge volume by eye on a waveform or by ear alone. The result is a clip that feels quiet next to music, or dialogue that shouts over the score. Loudness can be measured, and the standard measurement is LUFS. This guide explains what the numbers mean and how to use them in an ordinary edit, without needing a mastering studio.

Peak level and dBFS

Digital audio is measured in decibels relative to full scale, written dBFS. Zero dBFS is the highest level the format can store. Every value below zero is quieter. A signal that touches 0 dBFS is clipped, which means the top of each wave is cut flat. Clipping adds distortion that cannot be undone later, so peak level is the first thing to protect.

Peak meters show the loudest single sample in a short window. They are good at warning you about clipping, but poor at predicting how loud something sounds. The ear responds to energy spread over time, not to one sample, so a clip with sharp spikes can sound quieter than a clip with a flatter, denser waveform even when its peak reading is higher.

True peak

Sample peaks can miss overs. When a file is converted back to analog or resampled, the reconstructed wave can rise above the highest stored sample. True peak metering oversamples the signal to catch these inter-sample peaks. A common safe ceiling for delivery is -1 dBTP, which leaves a small margin for the codec and the playback device.

What LUFS measures

LUFS stands for Loudness Units relative to Full Scale. The method is defined in ITU-R BS.1770, and the European broadcast recommendation EBU R 128 builds on it. The meter applies a frequency weighting that roughly follows human hearing, then averages the energy over time. The output tracks perceived loudness far better than peak level does.

A loudness meter usually shows three readings:

Momentary loudness averages over 400 milliseconds. It moves quickly and is useful for spotting sudden jumps, such as a door slam or a music sting.

Short-term loudness averages over 3 seconds. It is the best reading for judging a single scene or a single cut.

Integrated loudness averages the whole program. It is the number that delivery specifications usually ask for.

US broadcast documents sometimes write LKFS instead of LUFS. The two units are equivalent, so a target of -24 LKFS and a target of -24 LUFS mean the same thing.

Common targets

Each destination sets its own integrated target. The numbers below are the ones most often cited in published guidance. Always check the current documentation of the platform you deliver to, because specifications change.

Destination	Integrated target	True peak ceiling
EBU R 128 broadcast (Europe)	-23 LUFS	-1 dBTP
ATSC A/85 broadcast (United States)	-24 LKFS	-2 dBTP
Many music and video streaming services	about -14 LUFS	-1 dBTP

Streaming services normalize playback, which means they turn tracks up or down toward their own reference level. A master that is already quiet is not penalized for being quiet. A master that is very loud is turned down, and heavy limiting can leave it sounding flatter than it needed to. The practical lesson is to aim near the platform target instead of chasing the loudest possible file.

Metering in a real edit

Put a loudness meter on the master bus, then play the whole timeline from start to finish. Read the integrated value at the end. Do not stop the playhead and read a single momentary value, because that number tells you almost nothing about the full program.

Work in small steps. Change one thing, play through again, and read the result. Large jumps in level usually mean a clip has been pushed too hard, not that the target is wrong.

Dialogue-heavy videos often measure lower than music-heavy videos at the same perceived volume. Balance the relationship between clips first, so the voice sits at a consistent level from one scene to the next. Then set the master to the target. Balancing the master first hides inconsistent clips underneath a single average.

Gain staging comes before loudness

Loudness problems are usually gain problems in disguise. A reliable order of operations looks like this:

Clean each clip first, removing noise and rumble. Set clip gain so dialogue averages around a comfortable level. Use bus processing, such as light compression on voice and gentle EQ, to even out the dynamics. Only then set the final loudness on the master.

Starting at the end of this list is a common mistake. A limiter on the master can make a badly balanced edit measure correctly, while the scenes still jump in volume when a viewer watches. The meter confirms the average; your ears confirm the consistency.

Limiters are the last stop

A limiter sets a ceiling. It lets you raise integrated loudness until the peaks reach that ceiling and then stops them from going any higher. A few decibels of gain reduction at the loudest moments is usually transparent. Heavy limiting flattens transients, which are the sharp starts of consonants and percussion hits, and makes speech sound pumped or tired.

Set the limiter ceiling to the true peak target for your destination, then raise the input until the integrated reading reaches the loudness target. If you need more than about six decibels of reduction to hit the number, the mix is probably too dynamic for that target. Fix the balance first, or accept a slightly lower target.

Check the export, not just the timeline

The timeline can sound different from the exported file. Codecs add their own processing, and some players apply their own normalization. Export a short test clip with the real delivery settings, then measure the file itself with a meter that reports integrated LUFS and true peak.

Listen on headphones and on speakers. Compare the result with a finished video you admire on the same platform. Your ears are the final judge of balance, while the meter is the reference that keeps your decisions repeatable from one project to the next.

Key takeaways

LUFS measures how loud something sounds over time, while dBFS and true peak protect you from clipping. Choose the target that matches the place where people will watch or listen. Meter the whole program rather than a single moment, balance clips before touching the master, and treat the limiter as a ceiling rather than a volume knob. A consistent edit that measures close to the target will beat a louder file that jumps around between scenes.