Six processes, five controls. The enhancer runs a recording through the chain a mixing engineer would build by hand, and asks you for the amount rather than for twenty-four separate numbers.
What runs, and in what order
The order is not arbitrary. Each stage cleans up the input the next one has to make a decision about.
- High-pass. Removes handling noise, desk thumps, and air conditioning below the voice. Doing this first means the compressor is not reacting to energy nobody can hear.
- Denoiser. Tracks the noise floor in twenty-eight bands and holds down whatever stays at floor level. Hiss, room tone, and fan noise.
- De-esser. Catches sibilance in the band above 6 kHz. It runs before the tone stage, because lifting the top of a sibilant recording makes the problem worse.
- Tilt EQ. Trades low-mid weight for presence around a pivot near 900 Hz.
- Compressor. Brings the quiet parts closer to the loud ones so the file holds up on a phone speaker.
- Limiter. Enforces the ceiling. Nothing crosses it whatever the stages before it did.
Each of those is a full tool on this site in its own right. The enhancer drives the same code, so a file processed here and a file processed by hand through the six pages arrive at the same place.
Strength
Strength is the master amount, and everything except the ceiling scales with it.
At 0 the chain is close to transparent: the high-pass drops to 20 Hz, the denoiser reduces nothing, the de-esser has no range to work with, the tilt is flat, and the compressor sits at 1:1. Turn it up and every stage moves together toward the full amount its material profile allows.
Start at 60 and listen. On a decent recording, anything above 80 usually reads as processed rather than better.
The three trims
Noise reduction, Clarity, and Levelling are shares of Strength, which is why they start at 100 rather than in the middle. A trim exists to take one part of the treatment back out.
- Noise reduction to 0 on a clean studio take. The denoiser has nothing to remove and its band gating can only cost you detail.
- Clarity to 0 on something already bright. The tilt stops lifting the top.
- Levelling to 0 on music you intend to master properly later. The compressor goes to 1:1 and the dynamics survive.
Material
The three material settings change the frequencies and the ratios, not the structure.
Voice high-passes at 85 Hz, de-esses at 6.5 kHz, pivots the tilt at 900 Hz, and compresses at up to 4:1. It assumes one person talking.
Music high-passes at 30 Hz so the bass survives, de-esses higher and lighter, pivots at 1.2 kHz, and compresses at up to 2.6:1 with a slower attack so transients keep their front edge.
Field recording is the aggressive one: 110 Hz high-pass, up to 26 dB of noise reduction, a 6.5 dB tilt, and up to 5:1 compression. It is built for phone captures, lecture halls, and anything recorded next to a road.
Ceiling
The ceiling is the one control Strength does not scale. Whatever the rest of the chain does, the limiter holds the output below this number.
Leave it at -1 dB for most things. Drop it to -3 dB if the file is going to be re-encoded to MP3 or AAC afterwards, because lossy encoding can overshoot the peaks it was given by a decibel or more.
Where it stops
An enhancer cannot recover what was never recorded. A clipped file needs the declipper before it gets here, a file recorded in a stairwell needs the dereverb, and a hum at 50 or 60 Hz is better taken out by the hum remover, which targets the harmonics rather than shelving the whole low end.
It also does not measure the whole file. If you need to land on -14 LUFS for a streaming platform, run the normalizer after this.