LeCoqRoti
Sound designers and mastering engineers often reach for SpectraLayers when a waveform alone hides the problem. Here are the first things to try.
3D view — seeing the mix
Rotate the spectrum into three dimensions. A click lets you inspect resonant peaks from an angle no equaliser can show you.
Unmix — stems from a single file
The AI-driven processes let you treat a finished mix as if it were separate tracks. Pull a vocal forward, isolate a reverb tail, or silence an instrument.
Pattern Finder — similar events, one action
Mark a squeak, a plosive, or a chair scrape, then ask Pattern Finder to locate every similar sound event in the spectrum. Edit them all at once.
Recent changes: what 13.0.20 refines in daily work
The installer release notes for version 13.0.20 are concise, but a few adjustments surface quickly in practice. Layer handling feels more predictable when you move selections between the Layers panel and the main display: selection edges snap with less drift, which matters when you are isolating a short transient and the slightest offset pushes it into the next spectral bin. The unmixing engine has received a quiet tune-up, too. Separation of tightly overlapping elements — a vocal riding the same mid‑range pocket as a piano, for instance — produces fewer of the watery artefacts that earlier builds could leave behind on sustained notes.
ARA 2 communication with Nuendo and Cubase is noticeably stabler. Round‑trips that previously required a manual resync after heavy editing now hold their alignment, saving a small but persistent annoyance. On Apple Silicon machines the 3D rendering pipeline uses the GPU more efficiently; spinning a dense spectral view no longer introduces the micro‑stutter that could interrupt close inspection of a high‑frequency band. For users who rely on the Voice Denoiser as a first pass on dialogue, the threshold response has been tightened — it now engages more gradually, which helps preserve breath detail that earlier versions sometimes trimmed too aggressively.
Spectral selection depth: beyond Adobe Audition’s spot brushes
Adobe Audition: Audition includes a spectral frequency display and a handful of spot‑healing brushes, and for broadcast editors who already live inside Creative Cloud the convenience is real. Where SpectraLayers pulls ahead is in the depth of its spectral selection model. Audition treats the spectral view largely as a diagnostic window with repair tools attached; SpectraLayers treats every pixel of the spectrogram as editable material. Unmixing stems, isolating harmonics, and running Pattern Finder across a whole file are operations with no direct equivalent in Audition’s toolkit.
iZotope RX: RX is the obvious comparison and the more mature platform in several respects — its dialogue‑specific modules, loudness compliance tools, and batch processing are deeply refined. Yet RX’s spectral editing remains module‑based: you apply processes to selections, but you do not freely paint, clone, or move spectral material the way you do on a SpectraLayers canvas. For restorative work on a single problematic file, RX often feels faster; for creative re‑sculpting of audio, or for extracting elements to repurpose in a production, the direct‑manipulation model of SpectraLayers offers a different kind of control.
Accusonus ERA bundles (now part of Meta) offered simpler one‑knob solutions for de‑noising and de‑essing. They were effective for quick fixes but never competed as spectral editors. SpectraLayers sits at the opposite end of the spectrum: it demands more from the user and returns far more latitude in exchange.
Chair squeak at 2:14: isolating a transient in a finished mix
Consider a mastering engineer who receives a stereo mix with a crisp, forward vocal that the client loves — and an unintended chair squeak at 2:14, buried just under the second chorus. The squeak spans roughly 2.5 kHz to 4 kHz and overlaps the vocal’s presence range. A broadband cut would dull the voice; a narrow EQ notch would soften the squeak but leave its harmonics intact.
She opens the file in SpectraLayers, switches to 3D view, and rotates the spectrum until the squeak stands out as a thin vertical ridge. With the lasso tool she traces around it, feathering the selection boundary by two bins. A right‑click applies a gentle gain reduction — just enough to push the noise below the perceptual threshold. She then highlights the same frequency band across the entire track and runs Pattern Finder with a similarity threshold of 85%. Two other faint squeaks appear in the third chorus, which she treats identically. The vocal remains untouched because the selections never touched its fundamental or its strongest harmonics. The whole process takes under four minutes, and the client never needs to know the squeaks were there.
Brittle transients after unmixing: fixing metallic snare residue
Unmix artefacts on percussive transients. When a stem extraction leaves a brittle, metallic residue around snare hits or hi‑hats, the issue is usually that the AI assigned part of the transient to the wrong layer. Instead of re‑running the full unmix, isolate the affected moment in the Layers panel, copy the transient from the original mix layer, and paste it directly onto the stem. Blend the edges with a short crossfade — the manual override often sounds cleaner than a second AI pass.
3D view becomes sluggish on large files. Long recordings at high sample rates can make the 3D render stutter, even on capable hardware. Reducing the FFT size temporarily to 2048 or 1024 samples lightens the load for navigation; you can restore a higher resolution once you zoom in on a specific region.
Selection drift after zooming. If a carefully drawn selection shifts by a few pixels when you zoom in for detail work, check that snapping is enabled and set to a sensible grid. Disabling "Snap to zero crossings" can also prevent the selection boundary from jumping to an unintended point when the waveform is asymmetrical.
Voice Denoiser over‑suppresses soft syllables. On a quiet passage where the denoiser flattens the ends of words, reduce the threshold and increase the attack time slightly — around 15 to 20 milliseconds — so the processor lets through the natural decay of each syllable. A second, gentler pass often yields better results than one aggressive sweep.
Phone buzz between syllables: why dialogue editors zoom in
The tool’s design assumes patience and a willingness to zoom in. Post‑production dialogue editors who need to remove a phone buzz from a single word without re‑recording will find it indispensable, because the spectral brush can erase the buzz between the formants of the voice. Forensic audio examiners, who work with recordings where every syllable matters, use the same precision to clarify speech buried under mechanical hum. Music producers who treat samples as raw material rather than fixed assets — slicing out a guitar riff from a crowded mix, or lifting a drum break cleanly — gain a library of reusable parts that no EQ‑based extraction could provide. Mastering engineers, as in the scenario above, use it for surgical corrections that leave the musical balance intact.
What links these users is a need to edit inside the sound rather than across it. If your work rarely goes deeper than fades and broad‑stroke equalisation, the learning curve may not justify itself. If you regularly wish you could reach into a recording and move just one element, SpectraLayers is built for that impulse.
Correspondence 0
Take a chair or join the table to send a letter.
No correspondence on this note yet. Letters are read before they appear.