Why Your Slowed Edit Clips: Headroom, True Peak, and the Numbers That Decide
Data engineer who loves building high-performance data and web-related tools. Creator of SlowedReverbMaker.net, implementing browser-side digital signal processing (DSP) to democratize audio editing.
There is a specific kind of disappointment in finishing a slowed edit, playing it back, and hearing the low end crackle. The instinct is to blame the effect — too much reverb, too much bass, the tool is bad. Almost always, none of those is the culprit.
The edit distorted because the file ran out of room, and nothing in the process told you how much room there was to begin with. That quantity has a name, it is measurable in about a second, and once you can read it the whole problem becomes arithmetic instead of guesswork.
Everything below is measured. The figures come from running the site's own analysis code over synthetic signals, which means you can reproduce them rather than take my word for it.
Try this while you read: Volume Booster
Make audio louder without clipping. Free, in your browser, no signup.
1. The Vocabulary You Need
Four terms carry the whole explanation. Getting them straight is most of the work.
- 0 dBFS
- Digital full scale — the largest value the format can represent. Unlike analogue, there is nothing above it. It is a wall, not a guideline.
- Clipping
- What happens when a signal would exceed the ceiling. The system cannot represent the value, so it substitutes the largest one it has, and the top of the waveform is replaced with a flat line. That flat line is the crackle you hear.
- Headroom
- The gap between a file's current peak and 0 dBFS. It is a property of your specific file, and it is the only quantity that decides whether a boost will survive.
- Boosting
- Addition, not redistribution. Turning the bass up does not borrow level from the midrange — it adds level, and the sum has to fit under the ceiling.
Why Digital Fails Differently From Tape
Analogue tape pushed past its limit compresses gracefully and adds harmonics that people pay money to emulate. Digital does no such thing. There is no soft region approaching the ceiling and no pleasant saturation past it — the waveform is simply cut off.
This is why 'a bit too loud' is a real category in analogue and essentially does not exist in digital. You are either under the ceiling or you are destroying the signal.
2. A Measured Example
Here is the situation in its simplest form, using real numbers from this site's own analyser rather than a hypothetical.
What That Result Means
There is no such thing as +5.39 dBFS in a finished file. Everything above zero is removed on export, so roughly five and a half decibels of the waveform gets flattened against the ceiling.
The tool did exactly what it was asked. The request was impossible, and the only way to know that in advance was to look at the headroom first.
This is why '+6 dB of bass' is not a meaningful recommendation on its own. On a quiet, dynamic recording peaking at −12 dBFS, +6 dB is comfortable. On the file above it is 5.39 dB of destruction. Same setting, opposite outcome, and the difference is entirely a property of the source.
3. The Peak That Is Not in the File
Now the part that catches out people who thought they were being careful. You can check every sample in a file, find that none exceeds full scale, and still get distortion on playback.
- Sample peak
- The highest individual sample value in the file. What a simple peak meter reports.
- True peak (inter-sample peak)
- The highest value of the reconstructed waveform between samples. Samples are measurements taken at intervals; the real signal is the smooth curve through them, and that curve can go higher than any point defining it.
Seeing It in Action
This is easiest to demonstrate with a deliberately awkward signal: a tone at exactly a quarter of the sample rate, positioned so the samples land symmetrically either side of each crest rather than on top of it.
Why This Is Not Pedantry
Three decibels is a lot to be wrong by. A file that looks like it has 3 dB of headroom may have none at all, and the overshoot only appears at the moment of playback — which is to say on someone else's device, after you have published it.
This is the reason broadcast loudness standards specify a true-peak ceiling rather than a sample-peak one, and why mastering engineers leave margin below zero rather than pushing right up to it. This site's volume booster estimates true peak by reconstructing the waveform at four times the rate with a windowed-sinc interpolator, which is why it reports a more pessimistic number than a simple peak meter.
4. Loud Is Not the Same as Peaky
There is a second measurement that matters, and conflating it with the first is where a lot of confusion starts.
- Peak
- The single highest instantaneous value. Tells you how close you are to clipping.
- LUFS (integrated loudness)
- How loud something actually sounds to a person. Measured by filtering the audio to approximate the ear's frequency response before measuring energy, because human hearing is not equally sensitive across the spectrum.
The Same Peak, Very Different Loudness
Three pure tones, all normalised to exactly the same peak of 0 dBFS, measured for integrated loudness:
- 1 kHz sine — −3.3 LUFS
- 40 Hz sine — −9.3 LUFS
- 10 kHz sine — +0.3 LUFS
Why This Is Exactly the Trap
Identical peaks, and yet the 40 Hz tone measures six decibels quieter than the 1 kHz one. Your ear does not weight deep bass as heavily as the midrange, and the loudness standard encodes that.
Now combine it with the previous section. Bass boosting adds a great deal of peak level while adding much less perceived loudness. You push the low end up, it does not sound dramatically louder, so you push further — consuming headroom at a rate the result does not seem to justify.
The edit that clips is very often one where someone kept boosting precisely because it did not sound loud enough yet.
5. Why Slowing Makes All of This Worse
Slowed edits are unusually prone to this for reasons that compound.
Slowing with varispeed lowers every frequency in the recording. Content that used to sit in the midrange moves down toward the bass, which means more of the track's energy ends up in the region your ear discounts — and, if you are also running a bass boost, more of it falls under the shelf's corner and gets lifted. The edit gets heavier and less obviously loud at the same time, which is exactly the condition that invites over-boosting.
Reverb then adds its own contribution. A reverb tail is additional signal summed on top of the dry track, so a wet mix does not merely place the sound in a room, it raises the level. Where the tail of one note overlaps the attack of the next — which is most of the time at longer decay settings — those overlaps add, and the peaks land higher than either sound alone.
None of this is a reason to avoid the effects. It is a reason to set them in a sensible order: get the speed and reverb where you want them, then look at the headroom, then decide how much bass the file can actually take.
6. What To Actually Do
The workflow that avoids all of the above is short.
- Measure before you boost. Run the file through the volume booster and read the headroom figure. That number is your budget for everything that follows.
- Treat the budget as real. If a file has 3 dB of headroom, a +3 dB shelf is roughly where the loudest moments start hitting the ceiling. Do not spend more than you have.
- If there is no headroom, reduce first and boost afterwards. Boosting into a wall and then turning the clipped result down does not undo the damage — the flattened peaks are already baked in.
- Leave extra margin for MP3. The reconstructed waveform can exceed the sample values, so a file that measures safe can still overshoot on playback.
- Judge bass on headphones or a speaker with a real low-frequency driver. A laptop or phone speaker cannot reproduce the range you are adjusting, so you will reliably push it too far trying to hear something the hardware was never going to produce.
- Export WAV when anything downstream will encode again. It costs disk space and removes an entire generation of lossy processing.
The Short Version
Distortion in a slowed edit is a budgeting failure, not an effects failure. There is a fixed amount of room above your track's current peak, boosting spends it, and the amount available is a property of your specific file that no general recommendation can know.
Measure it once and the guesswork disappears. Try it on your own track with the volume booster, or go straight to a bass boost now that you know what the number means.
Method
Every figure in this post was produced by running this site's `analyzeLoudness` and `renderProcessed` functions over synthetic test signals in a Node environment with a Web Audio implementation, rather than being quoted from elsewhere. True peak is a 4x windowed-sinc estimate; integrated loudness follows ITU-R BS.1770 with the standard two-stage gate. The inter-sample example uses a cosine at exactly one quarter of the sample rate with a 45-degree phase offset, which is the classic worst case and produces a 3.01 dB overshoot analytically as well as by measurement.