Why does the first ~1 second go silent on a freeze / bounce?

Standard mode reports 500 ms latency for host PDC. During Live's freeze the host signals offline mid-stream, and the worker thread re-primes for one chunk. Output is correct from the first audible frame onward.

Diff mode sounds different on each play. Bug?

Phase of the residual depends on chunk-grid alignment — same SR + same transport start gives bit-identical phase. Spectrum and RMS are stable across runs; only the per-sample residual rotates.

The result sounds dull. What knob fixes that?

Lower Aggressiveness, or switch to HPSS harmonic focus (preserves full bandwidth and only cleans the percussive split). Pulling the high-shelf emphasis to −6 dB also tells the model to leave the top end alone.

I'm getting dropouts on an M1 8 GB.

Use Standard mode (ECO is tighter). Reduce HPSS kernel from 47 → 31 → 15. Trim other plug-ins on the chain — total plug-in latency feeds back into the host's buffer math.

Is there a free trial?

Yes — 14 days of full quality with no gaps, every preset, every parameter. The same installer becomes the licensed build the moment you drop in a key. After the 14-day trial the plug-in inserts a brief silent gap every 60 seconds until you activate.

de-artifact — AI Music Artifact Removal

Try it free for 14 days, buy when you're sold

14-day full trial — every preset, every parameter, no gaps. After 14 days the plug-in inserts a brief silent gap every 60 seconds until you activate.

Start the 14-day trial

14-day free trial

Full quality, every preset, every parameter, no gaps for 14 days. After that the plug-in inserts a brief silent gap every 60 seconds until you activate with a license key.

de-artifact-0.1.8.pkg macOS Universal · 200 MB de-artifact-0.1.8-setup.exe Windows installer · 38 MB de-artifact-0.1.8.zip Windows portable · 68 MB

macOS — Apple Silicon + Intel · Windows — x86_64 · Linux — roadmap

Buy de-artifact

Launch price — ends Jul 31

One-time purchase, three machines, free updates within v1.x. The license key arrives by email and unlocks the plug-in instantly — whether you're on day 1 of the trial or past day 14.

$179$129one-time

Try online →

Hear it on real tracks

Pick a track and tap Original / Cleaned / Diff to A/B in place. Playback is synchronized across all three — no re-buffer, no click, no time jump. Diff plays only what de-artifact removed.

0:00 0:00

Keyboard: 1 Original · 2 Cleaned · 3 or D Diff · M Mix/Vocal · Space play · ←/→ ±5 s

Tip: if Diff sounds like music, the preset is too aggressive — back off Aggressiveness in the plug-in. Mix is the full bounce; Vocal is the htdemucs-separated vocal stem.

Try it on your own audio

Drop any AI-generated audio. de-artifact runs on GPU in the cloud and returns a 30-second A/B preview — no account needed.

WAV / MP3 / FLAC · max 50 MB · first run may take ~60s to warm up the GPU

🎵

Drop your audio file here

or click to browse

Processing

Submitting to GPU…

Preview ready

First 30 seconds · AAC 128k

Original

7 factory presets

Each preset is tuned on a real corpus of separated stems and AI-generated tracks.

Preset	Use it for	Character
HPSS focus q=0.6 · 12 dB default	Starting point for any source. Balanced cut.	Listener-validated baseline
HPSS focus q=0.6 · 20 dB	Fragile material — softer transitions	Smoother knee than 12 dB
HPSS harmonic focus	Vocals, piano, melody — preserve harmonics, clean percussive only	−7.7 dB GR, RMS unchanged
HPSS balanced	Mid-tier, between conservative and AI-aggressive	Light overall touch
Lossy conservative	MP3 / Opus 128 kbps — codec residual hash	Most subtle — barely audible
AI aggressive	Suno / Udio / Stable Audio full mix or stem	HPSS kernel 47 — broadband cleanup
AI vocal stem	htdemucs / Spleeter separated vocal stems	Strongest preset — kernel 15, harmonic boost 1.5

Features

Two latency modes

Standard 500 ms for mixing — biggest RT margin, highest quality. ECO 186 ms for tracking and live monitoring. Toggle in the corner of the spectrum view.

4-band emphasis EQ

Drag spectrum handles to tell the model where to focus. The boost is fed only to the model — listener audio is auto-inverse-EQ'd, so tone is preserved.

Three monitor toggles

Bypass for A/B with PDC preserved. Diff to hear exactly what's being subtracted. De-ess solo to isolate the sibilance side-chain.

Sidechain de-esser

Built-in HPF / BPF de-esser applied after the model's pass — clean up residual sibilance without re-tuning the main attenuation.

3 spectral algorithms

SoftSub (default, power-domain), SpecSub (linear, more aggressive), Wiener (SNR-based for noisy sources). One click to compare.

Standalone app included

No DAW required. Route through any Core Audio interface — system SR / period size configurable. Same engine as the plug-in.

How it works

de-artifact runs three stages in series. The audio path stays the model's domain until the very end — clean residual subtraction with no manual EQ tricks.

# Per-frame, real-time
audio (44.1k) → STFT → HPSS (median filter) ┬─ H_mag → ArtifactUNet → mask_H → soft_sub
                                       └─ P_mag → ArtifactUNet → mask_P → soft_sub
                                                                       ↓
                              clean = H_clean + P_clean → ISTFT → output

HPSS splits the spectrum into tonal and transient components. Each is masked independently by ArtifactUNet, a 4.2M-parameter forensic network trained on AI-generated music. The masks drive adaptive spectral subtraction — energy below the noise floor is removed, music above it is preserved.

System requirements

ArtifactUNet (4.2M params) runs on the audio thread alongside HPSS analysis. Standard mode keeps a generous RT margin even on entry-level Apple Silicon.

Requirement	Minimum	Recommended
macOS	Universal Binary — Apple Silicon (arm64) + Intel (x86_64) in one .pkg · macOS 11 (Big Sur) or later
Windows	Windows 10 x64 or later · .exe installer · VST3 + Standalone
CPU	Apple M1 / Intel Core i5 (4-core, 2.0 GHz+) / AMD Ryzen 5	Apple M2 or later / Intel Core i7 / AMD Ryzen 7
RAM	8 GB (Standard mode only; ECO may dropout on stereo)	16 GB+ (ECO + multi-instance comfortable)
Disk space	~150 MB (plug-in + standalone + ONNX models)	—
Audio buffer	≥ 256 samples (Standard) / ≥ 512 samples (ECO)	1024 samples for tracking
Sample rate	Any — rubato resampler 8 k–384 k. Internal model runs at 44.1 kHz.
Hosts (macOS)	VST3: Live, Reaper, Bitwig, Studio One · AU: Logic Pro, GarageBand, MainStage · CLAP: Reaper, Bitwig · Standalone (Core Audio)
Hosts (Windows)	VST3: Live, Reaper, Bitwig, Studio One, FL Studio · Standalone (WASAPI / ASIO)
Logic / Pro Tools	AU supported. AAX (Pro Tools) on the roadmap.
Linux	Roadmap.

FAQ

Why does the first ~1 second go silent on a freeze / bounce?: Standard mode reports 500 ms latency for host PDC. During Live's freeze the host signals offline mid-stream, and the worker thread re-primes for one chunk. Output is correct from the first audible frame onward.
Diff mode sounds different on each play. Bug?: Phase of the residual depends on chunk-grid alignment — same SR + same transport start gives bit-identical phase. Spectrum and RMS are stable across runs; only the per-sample residual rotates.
The result sounds dull. What knob fixes that?: Lower Aggressiveness, or switch to HPSS harmonic focus (preserves full bandwidth and only cleans the percussive split). Pulling the high-shelf emphasis to −6 dB also tells the model to leave the top end alone.
I'm getting dropouts on an M1 8 GB.: Use Standard mode (ECO is tighter). Reduce HPSS kernel from 47 → 31 → 15. Trim other plug-ins on the chain — total plug-in latency feeds back into the host's buffer math.
Is there a free trial?: Yes — 14 days of full quality with no gaps, every preset, every parameter. The same installer becomes the licensed build the moment you drop in a key. After the 14-day trial the plug-in inserts a brief silent gap every 60 seconds until you activate.

de-artifact