de-artifact 0.4 Walkthrough: What to Listen For, Chapter by Chapter

By IntrectOctober 5, 2026Last reviewed Tutorialde-artifact

This is a written companion to our 17-minute de-artifact 0.4 walkthrough. The video runs the plug-in on full mixes, separated stems and older digital samples, alternating bypass and processed audio and then playing the residual — the part the plug-in removes. Below is each section, with what to listen for and links that jump straight to it.

How to watch it

Use headphones and listen in three passes: bypass, processed, then residual. The residual is the honest test — it should sound like noise and grit, not like the song. The video was recorded with de-artifact 0.4.2; the presets it uses ship as factory presets from 0.4.3.

The full walkthrough, 17:39. Watch on YouTube

Chapters

Breaks and older digital samples (0:18–6:06)

Listen to the transients first. Each loop alternates bypass and processed audio, then plays the residual. The question is whether the attack of the hit survives while the grainy, smeared layer around it goes.

How the mask works (6:06–7:02)

A short explanation of what the plug-in estimates: per time–frequency bin, how much of the energy is residue to reduce. It is a mask on the existing signal, which is why the residual monitor is a fair test of what is being taken away.

Full mixes (7:02–9:50)

On a finished mix, listen to vocals and cymbals together. A good setting removes the chirping and watery shimmer without making the mix duller; if the residual starts to carry melody or lyrics, the setting is too strong for that track.

High-frequency reconstruction (9:50–11:14)

The 0.4 series can regenerate the upper band from the cleaned signal (the Brightness control). The controlled example low-pass filters the input at 6 kHz and compares Brightness 0 with +1. The generated band is an estimate, not the lost original restored.

More full mixes (11:14–13:05)

Two more starting points for whole-mix cleanup, including the preset tuned on Suno v6 output.

Processing time and latency (13:05–14:19)

Why the plug-in reports latency to the host, and what the same-machine Core ML comparison measures. The reported delay is compensated by the DAW, so A/B with Bypass stays aligned.

Separated stems (14:19–17:23)

Stems from a separator carry their own residue: bleed and smeared edges. Here the presets are applied stem by stem.

Try the same thing on your own audio

Results depend on the source, so the useful test is your own track. The free online preview processes up to 30 seconds of one file so you can compare original and processed in the browser. In your DAW, the plug-in has a 14-day free trial.

Hear it on your audio

Free 30-second preview in the browser, or the full plug-in for 14 days. Code INTRECT10 takes $10 off a license.

Related guides