Start with the problem you can hear. These resources help you choose a cleanup workflow, learn the controls and decide whether the result is worth keeping. de-artifact is an audio restoration plug-in, not a stem separator or an AI-music detector.
Start here
- Read the de-artifact manual for installation, presets and exact control behavior.
- Try the online preview on a short excerpt before installing.
- Get the plug-in and start the trial from the official product page; check the current host and system requirements there.
- Listen to the existing before/after demonstrations, then repeat the comparison on your own material.
- Follow the de-artifact 0.4 walkthrough chapter by chapter — a chapter guide to the 17-minute walkthrough: breaks, full mixes, stems, high-frequency reconstruction, masking and latency, and what to hear.
Match the symptom to the workflow
Watery sustains, metallic shimmer or smeared detail
For residue within a generated track or separated stem, test de-artifact on a copy. Follow the AI-music cleanup guide and compare a modest setting with Bypass. Not every defect is removable; processing cannot recover irreversibly lost information.
Recognizable drums or accompaniment in the wrong stem
This is stem bleed, not just a damaged texture. See de-leak-rt and the bleed-removal workflow. If both bleed and residue are present, assess them separately rather than increasing artifact reduction to suppress an instrument.
One narrow ringing resonance
First check whether ordinary EQ or a dedicated resonance suppressor such as ENZYME addresses the audible peak. Two guides cover that narrow case in detail: killing snare ring without losing the crack, and removing room resonance from vocals. The de-artifact resonance controls are explained in the manual. Do not apply stronger broadband cleanup solely because one note rings.
A question about whether the music is AI-generated
That is a detection question: use the ArtifactNet information and analysis workflow. Audio quality and detector scores are separate questions. A cleaner sound is not proof of human authorship, and cleanup does not promise a lower detection score.
A level-matched A/B listening checklist
- Preserve the original file unchanged. Work on a duplicate track or a lossless working copy, and keep the unprocessed export.
- Loop the same short passage, including a sustained note and a transient. Start with a suitable preset and conservative Strength.
- Compare processed audio with the plug-in's Bypass, which preserves latency alignment. Use your DAW's metering and gain adjustment on comparison copies to match loudness; louder is not automatically better.
- Enable Residual to audition what is being removed, then turn it off before judging the cleaned output. If vocal body, useful harmonics or drum transients are being lost, reduce Strength.
- Check quiet and dense passages, both solo and in the mix. Listen for softened consonants, weakened attacks and altered tone, not just less fizz.
- Keep the processed version only if the level-matched comparison improves the audible problem without an unacceptable trade-off. Record the preset and settings so you can repeat it.
- Evaluate audio quality separately from detector scores. Do not use a detector-score change as the success criterion for restoration.
Guides for your source
Hear eight cleanup tools on the same audio: a dated, matched-loudness comparison with methods, measurements and limitations. Intrect makes de-artifact, one of the tools tested; the measurements are not a sound-quality ranking.
- Suno artifacts — generated texture and export residue.
- Suno metal artifacts: fizzy cymbals and guitars — why Suno metal tracks turn fizzy and smeared, and how to clean distorted guitars, cymbals and screams without flattening the attack of the mix.
- Udio artifacts — assessing an existing track.
- Remove instrument bleed from a vocal stem — telling bleed apart from separation residue, and why a noise gate struggles.
- Clean a separated vocal stem — residue versus bleed.
- WAV versus MP3 — preserve the best available source and avoid repeated lossy encodes.
- What is an audio artifact? — the definition and the common types, before you choose a fix.
- Which tool removes AI music artifacts — a dedicated plug-in, an open-source script, a repair suite or EQ: what each targets, where it runs and its limits.
- de-artifact vs artefactFX — how the two tools differ in purpose, format and price, from public pages.
- Why AI sound effects sound metallic or synthetic — a sound-effects problem rather than a music one: what the sources say about the cause, and five general-purpose tools compared from their own pages. de-artifact has not been evaluated on sound effects.
- Stem bleed removal plug-ins compared — separation bleed versus microphone spill, and what four vendor pages say each tool is built for.
- Dynamic resonance suppressor plug-ins compared — when a suppressor fits, and what five tools state about formats, trials and latency.
- Harmonic percussive separation tools compared — HPSS is not stem separation: six plug-ins, scripts and libraries from their own pages.
- AI music restoration: what to run first — an ordered workflow that marks which steps our guides document and which are our default.
- Can you tell AI music from real music? — a 35-clip listening test with the answer after every clip, then a free preview of cleanup on your own track.
Writing about de-artifact? The press kit has product facts, official screenshots and third-party release coverage.
Need help with a specific setup? Visit support.