Best Audio Repair Plugins in 2026

Most of the repair work I do is not glamorous — podcast hosts beside air returns, rappers over laptop fans, VO artists in tiled bathrooms because it was the only quiet room. I'm YECK (Hamid Shekari), Mixing Engineer & Music Producer. I mix and restore dialogue weekly for clients and on the @yeckxo channel, mostly in Logic Pro 11 and Pro Tools 2025. This is not a ranked list from spec sheets — it is the eight tools I kept installed after running them on paid work over several months, with pricing and version notes re-checked May 27–28, 2026. See my author profile for background on how I mix and what I do not claim to test.

Method note (May 28, 2026): pricing and tier claims checked against vendor storefronts linked below. Hands-on: level-matched bypass listening on client and channel material — no blind panel, no null-sum shootouts, no Windows-only DAW pass, no long-form Atmos deliverable. I do not quote precise dB reduction or plugin dial positions in this guide unless they come from a manufacturer spec — listening notes below describe problem types and tradeoffs, not logged measurements.

Think in categories, not brands. Full suites (RX) when the damage is weird and you need a spectrogram. Live neural de-noisers (Clarity Vx Pro) when someone is still in the booth. Focused offline bundles (Acon for hum, clicks, clips; legacy Sonnox for buzz on archive jobs) when the problem has a name and you do not want to paint pixels. One-knob dialogue tools (ERA) when the producer is waiting. Cedar DNS One when location sound has to survive broadcast scrutiny. Gullfoss when the vocal is “noisy” but the meter says the room is fine — usually resonance, not noise. Mix those up and you will hate a perfectly good plugin.

For where repair sits inside the wider AI mixing stack, see the best AI mixing plugins in 2026. Full disclosure: I built MixingGPT, which helps with tool choice inside the DAW but does not process audio — it is last on this list for that reason. I receive no affiliate commission on RX, Waves, Cedar, or any other product here. Nothing below replaces a level-matched bypass check on your source material.

How I compared them

Everything was installed on one Mac Studio rig at 48 kHz in Logic Pro 11, with AAX behavior spot-checked in Pro Tools 2025 on client dialogue sessions — that is where iLok and format support actually matter for me. I did not run null-sum tests or a blind listening panel. This is practical, level-matched listening on material I actually work on: podcast archives, home-studio rap vocals, corporate lavalier dialogue, documentary location interviews, vinyl and cassette transfers, Zoom and phone-width sources, and mix sessions where a vocal sounded “noisy” but turned out to be moving midrange harshness, not room noise.

How deep the hands-on goes varies by tool. RX is years of paid dialogue and music restoration — RX 12 specifically since its April 28, 2026 launch, so v12 module notes reflect a few weeks of real jobs, not a year-in-review. Clarity, ERA, and Gullfoss are in weekly rotation on vocals and podcast work. Cedar DNS lives in a broadcast template I inherited; my Cedar-vs-RX notes come from limited exterior dialogue A/B on my rig, not a facility benchmark. Acon covers several one-off hum, click, and clip jobs. Sonnox appears only when old archive sessions load. MixingGPT examples come from product support and my own prompts — not a blind test against other advisors.

For each tool I noted whether consonants stayed intact after de-noise, whether room tone collapsed, whether clicks or hum returned on bypass, and whether the fix held after a small level change — the usual places algorithms lie.

Verified against primary sources (May 22–28, 2026): RX 12 tier names and module lists on iZotope's RX product page and RX edition comparison; Clarity Vx Pro list pricing on waves.com; Acon Restoration Suite 2; ERA bundles on accusonus.com; Cedar DNS One; Sonnox Restore retirement notice on sonnox.com; Gullfoss on soundtheory.com; MixingGPT plans on mixinggpt.com/pricing. Sale prices (Waves ~$50, iZotope seasonal cuts) were cross-checked against storefront banners the week of May 22–28 — they change; list prices in the table are anchors, not guarantees.

Not verified by me: every Waves sale window on every reseller, Cedar DNS pricing outside the UK storefront, long-form Atmos or broadcast multichannel workflows, Windows-only DAW behavior, or head-to-head null tests against Adobe Audition's repair suite, Steinberg SpectraLayers, Zynaptiq UNVEIL, or CrumplePop — competent tools I simply did not install for this pass. Where a section states a manufacturer feature (RX Spectral Recovery, Clarity Ambience Keeping, DNS Learn mode), it comes from the manual unless labeled as my session impression.

Competent tools left out on purpose

This list is what stayed on my Mac after paid work — not every repair product worth buying. Adobe Audition and SpectraLayers both do serious spectral repair if you already live in those apps. Zynaptiq UNVEIL is the usual answer when de-reverb — not de-noise — is the whole job. CrumplePop and similar podcast one-knob suites overlap ERA on speed. I did not A/B them here; if you own one of those already, compare against RX or Clarity on your material before buying another overlap.

Quick Comparison

Prices checked against vendor storefronts May 22–28, 2026. Reseller and sale pricing varies — confirm at checkout.

Where I'd start

• One purchase, every type of damage: RX 12 Standard (~$399 new). Budget time learning the editor — it pays back the first time you need spectral repair on a one-off intrusion, not just steady noise.
• Client in the room, bedroom vocal: Clarity Vx Pro on a Waves sale (~$50 if you catch it). Render RX later for mouth clicks.
• Under $100, obvious problems: Acon Restoration Suite 2 for hum, clicks, or clips. Add DeVerberate 3 separately if the bathroom VO is the actual issue.
• Podcast batch before coffee cools: ERA bundle — not theatrical, but fast.
• Wind on a doc interview: Cedar DNS One (~£249). Pair with RX Advanced when you still need de-click or spectral paint.
• “Fix this noisy vocal” but the noise is a whistle: Gullfoss before you buy another de-noiser.
• Not sure which bucket you are in: MixingGPT for a second opinion — it does not process audio.

1. iZotope RX 12 — The Full Restoration Suite

RX 12 shipped April 28, 2026 — rebuilt De-bleed and Breath Control, Stems View for working on separated dialogue or music stems, and the same spectral editor that still defines this category (per iZotope's RX 12 launch notes). Standard includes Voice De-noise, Spectral De-noise, De-hum, De-click, De-clip, De-reverb, Spectral Repair, Dialogue Isolate, and the standalone app. Advanced adds Spectral Recovery for Zoom-width sources, Scene Rebalance, Trim Silence, and other post-only modules iZotope keeps out of Standard.

The spectrogram is why RX stays installed even when I own several other tools here. I have used RX for years on dialogue and music restoration; RX 12 specifically since its April 28, 2026 launch, so v12 notes below mix long-standing RX habits with early v12 jobs — not a full year-in-review. Typical wins for me: steady HVAC hum on dialogue after Voice De-noise Learn, isolated bumps and intrusions in Spectral Repair, and stacked offline modules when one pass is not enough. None of that is fast, but nothing else on this list handles one-off spectrogram events as cleanly.

The tradeoff is time. Most modules are offline — render, listen, tweak, render again. Clarity Vx Pro is the better fit when the artist is still tracking. ERA beats RX on a quick podcast pass when the client is waiting and transparency is not the brief. List price is ~$399 for Standard new (~$169–$299 on upgrade paths), ~$1,399 for Advanced, Elements often ~$99–$129 before iZotope's seasonal sales. VST3, AU, AAX, and standalone.

2. Waves Clarity Vx Pro — Real-Time AI Vocal Cleanup

Clarity Vx Pro is the opposite end of the workflow from RX: one job, real time, no spectrogram. Analyze, set per-band reduction if you need it, listen while the talent is still in the booth. Waves calls the room-tone control Ambience Keeping — it strips fan noise and laptop whine without turning a bedroom vocal into a dead booth.

Home-studio and corporate sessions where I need to hear cleanup while tracking: Clarity is the obvious pick when RX feels like bringing a toolbox to hang a picture. On steady broadband noise I usually reach for Clarity before RX Voice De-noise; RX wins once clicks or reverb need fixing on the same file. It does not de-click, de-reverb, or isolate dialogue. Waves licensing is the other annoyance — skip the Update Plan for a year and you feel it at upgrade time. ~$249 list on waves.com, often ~$50 in sales; also bundled in Waves Ultimate. VST3, AU, AAX.

Fast Offline Fixes: Acon Digital and Accusonus ERA

When the problem has a label — hum, click, clip, plosive, Zoom reverb — and you do not need to paint a cough out of the spectrogram, these two bundles compete for the same slot. Acon is four serious offline modules for ~$99. ERA is one-knob speed with less surgical control. I keep both; which one opens first depends on whether I have five minutes or five seconds.

Acon Digital Restoration Suite 2

DeNoise, DeHum, DeClick, and DeClip — no de-reverb in the bundle. DeVerberate 3 is separate (~$99). I use DeHum on ground hum and wandering mains noise on bass and dialogue, DeClick on plosives and transient junk, and DeVerberate 3 — not the suite itself — when room tail is the main problem. Straightforward damage at ~$99 on acondigital.com — VST3, AU, AAX. No spectral paint, no dialogue isolation; a cough between words still wants RX.

Accusonus ERA Bundle

ERA is the “big knob, small brain required” set — De-Noiser, De-Verberate, Plosive Remover, De-Esser, De-Clipper, Noise Remover, Voice Leveler. On recurring podcast edits I lean on De-Noiser with Learn from silence between questions, Plosive Remover on hard consonants, and De-Verberate when remote interviews arrive with obvious room reverb. Fast enough to finish before a producer checks in — not what I would print for theatrical dialogue without comparing against Cedar DNS or RX Advanced on bypass. Pricing varies by tier and retailer; Accusonus/Meta storefront availability has shifted, so confirm the SKU before you buy. VST3, AU, AAX.

5. Cedar DNS One — When Location Sound Has to Print

Process and Learn. That is basically the UI — which still surprises people until they A/B it against generic broadband reduction on a windy doc interview. Cedar trained this algorithm on professional speech in noise, not laptop fan hum, so consonants and low-end body survive while wind, traffic, and rumble come down.

On exterior documentary dialogue I have A/B'd against RX Voice De-noise on my rig — DNS One kept consonants cleaner on wind and traffic noise on that material, though RX still wins on some indoor hum I have tried since. Messy exteriors are where DNS earns its slot in a template I inherited, not built. ~£249 + tax on cedaraudio.com (roughly $320 USD), 14-day trial, iLok, AAX/AU/VST3. It de-noises dialogue and nothing else — no de-click, de-reverb, or spectral paint. Full Cedar Studio bundles cost far more if you need the whole Cedar toolset.

6. Sonnox Restore — Legacy Oxford Clicks and Buzz (Retired)

Sonnox retired Restore in mid-2025 — no longer sold, support ended December 31, 2025. I am keeping it here because archive sessions still load DeClicker, DeBuzzer, and DeNoiser. DeClicker still helps on vinyl crackle and transient junk; DeBuzzer on steady buzz when the fundamental is obvious. If you do not already own it, RX De-click or Acon DeClick beats hunting dead stock (~$499 was the bundle list price). VST3, AU, AAX for existing licenses.

7. Soundtheory Gullfoss — Resonance, Not Noise

Gullfoss is the odd entry: dynamic EQ, not de-noise. Recover and Tilt, no bands, no Q — the algorithm hunts moving harshness and resonances. I reach for it when a vocal or drum bus has whistle-y or ringy buildup that moves with the performance, not when the room noise is actually high on the meter. ~$199, VST3/AU/AAX. Wrong tool for clicks, reverb, or broadband hum — right tool when static notch EQ keeps missing the peak because it shifts with the take.

8. MixingGPT — Pick the Tool, Not the Noise

MixingGPT loads in Logic, Ableton, Pro Tools, and the other major DAWs, but it does not process audio — it answers questions, accepts uploads, and suggests which repair category to try first. Typical prompts: whether a vocal needs de-noise or resonance control, whether wind-heavy dialogue wants Cedar-style de-noise or RX Voice De-noise, or what to listen for after a Clarity pass.

When the suggestion matches what I would have tried on bypass anyway, it saved setup time. When it does not, I still trust the final print to my ears — not the model. Free tier: 15 credits/week, text-only. Paid: Starter $9 (75 credits), Pro $19 (200 credits + audio/image analysis), Studio $49 (600 credits + Studio Drops). Needs internet; can be wrong on unusual source material. For the wider AI mixing picture, see best AI mixing plugins in 2026.

How to Choose Without Demo Fatigue

Match the plugin to the damage, not the marketing page. If you only remember three things: RX when the problem is weird, Clarity when someone is still performing, Cedar when location sound has to survive scrutiny. Everything else is a specialist — ERA and Acon for speed, Gullfoss when the vocal is harsh but not noisy, MixingGPT when you are staring at the file and cannot name the problem yet.

For specific chain-order questions — de-noise before de-ess, RX before compression — see guide to fixing muddy vocals.

Where Audio Repair Plugins Are Going Next

What I am seeing in 2026: live neural de-noise is normal — Clarity Vx Pro and RX's real-time Dialogue Isolate both run while talent tracks — a workflow that used to mean more offline rendering than it does now. Repair is also creeping into mixing sessions, not just post — Gullfoss on vocals, ERA on a quick podcast insert before the mix bus. The third shift is decision fatigue: more plugins, same broken recordings, and engineers using advisors to pick the first tool instead of opening four demos. For how that fits the wider mix workflow, see AI mixing vs traditional engineering.

Frequently Asked Questions

Short answers if you are skimming — the sections above have the workflow detail.

What is the best audio repair plugin in 2026?

There is no single winner. RX 12 Standard if you need spectral editing and a full module set. Clarity Vx Pro if you need live vocal cleanup without rendering. Cedar DNS if broadcast-grade dialogue de-noise on difficult location sound is the job. MixingGPT only if you need help choosing among those — it does not process audio.

What is the difference between spectral repair and AI-based noise reduction?

Spectral repair (like iZotope RX Spectral Editor) lets you manually paint away unwanted sounds in a frequency-vs-time display — you see the noise, select it, and delete it. AI-based noise reduction (like RX Voice De-noise or Waves Clarity) analyzes the audio and automatically separates the desired signal from noise using machine learning. Spectral repair gives you surgical control but is slow. AI noise reduction is fast but sometimes removes too much or leaves artifacts. Pro engineers use both: AI for the first pass, spectral for the fixes.

Should I use iZotope RX 12 or Waves Clarity Vx Pro for vocal cleanup?

Use Clarity Vx Pro when you need real-time monitoring during recording or a fast pass on broadband room noise in an untreated space — it runs live and Ambience Keeping keeps vocals natural. Use RX 12 when the problem goes beyond de-noise: mouth clicks, plosives, de-reverb, spectral surgery, or stacked offline modules across a full dialogue edit. A common pattern is Clarity for the first live pass and RX offline for clicks, reverb, and spectral fixes the real-time tool cannot handle.

How much do professional audio repair plugins cost in 2026?

Professional audio repair plugins range from roughly $99 to $1,399 at list price (verified May 2026 on manufacturer storefronts). iZotope RX 12 Standard is ~$399 new (~$169–$299 on upgrade paths). RX 12 Advanced is ~$1,399 new. Waves Clarity Vx Pro is ~$249 (frequently discounted to ~$50). Acon Digital Restoration Suite 2 is ~$99. Accusonus ERA bundles vary by tier and retailer availability. Cedar DNS One is ~£249 (~$320 USD). Sonnox Restore was ~$499 before retirement. Gullfoss is ~$199. MixingGPT has a free tier (15 credits/week) and paid plans from $9–$49/month. Sale prices change weekly — confirm before purchase.

Can AI audio repair replace manual restoration?

AI de-noise and spectral tools handle first-pass cleanup faster than manual editing used to on many dialogue and vocal sources. They still struggle with overlapping problems (noise plus reverb plus clipping on the same word), creative judgment about how much to remove, and surgical fixes that require painting in the spectrogram. A practical workflow is AI for the obvious noise pass, manual spectral repair for what the algorithm misses, and an advisor like MixingGPT when you are not sure which tool to reach for first.

What audio repair plugins do top engineers actually use in 2026?

RX and Cedar DNS show up constantly in post-house templates and dubbing workflows I inherit or hear discussed among dialogue editors. Clarity Vx Pro is common on home-studio and corporate vocal sessions that need live monitoring. ERA appears often in podcast and VO batches. Gullfoss gets used when harshness is mislabeled as noise. That is pattern recognition from paid work and peers — not a survey of every facility. MixingGPT is my product; it advises on tool choice only.

Corrections and updates: Fact-checked May 27–28, 2026 (RX 12 tiers, Acon Restoration Suite 2, Cedar DNS One, Sonnox retirement, MixingGPT plans). May 28 pass removed unlogged dB readings, dial percentages, and overly specific session vignettes; product facts and pricing remain tied to vendor sources above. If a storefront price changed after that date, trust the vendor — sale windows move faster than blog posts. Report a factual error via contact. For chain-order questions (de-noise before de-ess, RX before compression, etc.), see how to fix muddy vocals and best AI vocal plugins in 2026.