How to set levels for sound effects in video and UI demos

A whoosh that is too loud ruins a voice-over. A UI click that is too quiet vanishes. Most problems with sound effects are level problems, and most level problems come from mixing up three different numbers: the peak, the true peak and the loudness. This guide explains what each one tells you, how to use the measurements that come with a sound-effect library, and how to set levels in a video or a product demo without clipping.

Three numbers, three questions

Sample peak is the highest stored value in the file, in dB relative to full scale (dBFS). Zero is the ceiling: nothing can be stored above 0 dBFS in a digital file. It answers "how close is this file to clipping as stored?"

True peak, in dBTP, is an estimate of the highest level of the continuous signal that a player will reconstruct from the stored samples. Between two samples the reconstructed wave can rise above both, so the true peak can be higher than the sample peak. ITU-R Recommendation BS.1770 defines algorithms to measure both programme loudness and true-peak audio level, and the current version, BS.1770-5, was approved in November 2023.

Loudness, in LUFS (loudness units relative to full scale), is a measurement of perceived loudness over time, weighted to match how people hear. It answers "how loud will this feel?" The peak tells you about headroom, the loudness tells you about balance, and they are independent. A very short click can have a high peak and a low loudness. A quiet room tone can have a low peak and a loudness that matters over minutes.

Why sound effects are hard to measure

Loudness meters are designed for programmes, which means seconds or minutes of audio. A sound effect of a quarter of a second does not give a meter much to average, so its LUFS figure is a rough guide, not an exact one. That is why a good library gives you the measured numbers and still expects you to listen. Treat the loudness as a starting point and the peak as a hard limit.

Headroom: what you can add

Headroom is the distance between a file's true peak and the ceiling you choose. Broadcast and streaming ceilings differ. The EBU R 128 recommendation for broadcast sets a maximum true peak of -1 dBTP, a figure widely cited and used as a conservative ceiling in many workflows. We could not open the EBU document in our automated browser, which the EBU site blocks, so check the figure in the current text yourself. The FFmpeg documentation shows that targets vary by tool: its loudnorm filter defaults to a maximum true peak of -2.0 dBTP, an integrated loudness of -24.0 LUFS and a loudness range of 7.0 LU. There is no single correct number for every platform, so choose the ceiling for where the video will play, and keep a little more headroom than you think you need.

Gain in decibels multiplies the amplitude: a change of g dB multiplies the signal by 10 raised to g/20. So -6 dB multiplies by 0.501, which is close to half, and +6 dB multiplies by 1.995, about double. Adding gain moves the peak up by the same number of dB, so you can add at most (ceiling minus true peak) before you clip.

Worked example: how much can you add?

The measurements are invented. A manifest lists a gentle transition effect as -28 LUFS with a true peak of -6 dBTP. Your voice-over sits at -23 LUFS. You want the effect to feel about as loud as the voice at its loudest moment, which is rare, so you would first check by ear, but let us see what the numbers allow.

To match the voice by loudness you need -23 - (-28) = +5 dB.
Headroom to a -1 dBTP ceiling is -1 - (-6) = 5 dB.
So +5 dB just fits. The factor is 10 to the power 5/20, which is 1.778.

Now suppose the same effect had a true peak of -3 dBTP. Headroom is only 2 dB. You cannot add 5 dB without exceeding the ceiling. You have three choices: add 2 dB and accept that the effect is quieter than the voice, lower the voice, or use a limiter and accept that it changes the sound. This is why loudness and peak must be read together. It is also why many effects are supplied with a modest peak, so you can add gain, and why tidy fades matter: a click at the start of a file adds a peak you did not intend.

WAV or MP3, 48 kHz and 24 bits

Use WAV for editing and MP3 for quick previews and drops. WAV stores the audio without lossy compression, so you can edit and export it as often as you like. MP3 is lossy, and it adds a small delay at the start in some players, so keep it for situations where speed matters more than the last bit of quality.

48 kHz is the standard sample rate for video, so it avoids a conversion in most video editors. 24-bit audio has a theoretical dynamic range of about 144 dB, compared with about 96 dB for 16-bit, using the rule of thumb of 6.02 dB per bit. The practical benefit is that you can lower and raise levels in an editor with less risk of audible noise. Export your final mix at the delivery format your platform asks for.

A level workflow for video and UI demos

Set the voice or music first. Mix the lead element to the loudness you are aiming for, and leave the effects until later.
Choose effects by role. Transitions and whooshes support a cut. UI confirmations mark an action. Ambience sits below everything else. Impacts mark a moment.
Place before you adjust. Put the effect on the timeline and listen in context, at the volume your audience will use.
Use the manifest. Read the loudness and true peak for each file, and work out the headroom as above.
Keep effects below speech. An effect that competes with a voice-over costs the viewer attention. When in doubt, lower it.
Check on more than one device. Phone speakers, laptop speakers and headphones do not tell the same story.
Finish with a final measurement of the whole mix, and adjust the master, not the individual effects, if the total is off.

What not to assume

A peak limit does not guarantee a clean export. If you add gain, process and re-encode, the result can exceed the ceiling again. Measure the finished file.
Equal loudness is not the aim. A quiet ambience and a UI confirmation should not read as the same number.
A measurement is not a judgement. A file can pass every number and still sound wrong in context. Listen.
Platform rules change. Check the current specification of the service you deliver to.

Tool that fits

Clean Cut Sound FX (EUR 17, paid once) is a set of 100 sound effects made from scratch with our own code: 25 soft transitions, 20 digital UI confirmations, 20 whooshes, 15 gentle impacts, 10 ambience details and 10 abstract tonal accents. Each comes as a stereo WAV (48 kHz, 24-bit) and a 320 kbps MP3, with short fades at both ends, peaks kept at or below -1 dBTP and no clipping, as checked on every file. A contact-sheet PDF shows a waveform and duration for each, and a CSV manifest lists the measured loudness, peak and true peak. Loudness differs by group on purpose, so set levels in your editor. Unlimited personal and commercial projects; the files may not be resold or shared.

What to do next

Clean Cut Sound FX has 100 effects made from scratch: soft transitions, UI confirmations, whooshes, gentle impacts, ambience and tonal accents, each as a 48 kHz 24-bit WAV and a 320 kbps MP3, with a CSV of measured loudness and true peak. €17 once, for unlimited personal and commercial projects; the files may not be resold.

See the Sound FX

Questions people ask

What does dBTP mean?

It is decibels relative to full scale, measured as a true peak. A normal sample peak looks only at the stored sample values. A true-peak meter estimates the level of the signal the playback system will reconstruct between those samples, which can be higher. ITU-R BS.1770 defines the measurement algorithms for programme loudness and true-peak level.

Is WAV better than MP3 for sound effects?

For editing, WAV is the safer working format because it is not lossy. MP3 is lossy, so it is fine for quick drops into a timeline or for drafts, but you should avoid editing, re-saving and re-exporting MP3 repeatedly. If your project allows, export final audio from your editor, not from the effect files.

Should every effect have the same loudness?

No. A soft ambience bed and a sharp UI confirmation do different jobs, so equal loudness numbers would be wrong. Set levels by role: the voice-over or music leads, the effects support. Measure to find the headroom, then adjust by ear against the voice.

Sources

We checked each source on 17 October 2026. If a fact here is out of date, tell us at hello@hiddenway.info.