Build an iZotope RX 12 Module Chain That Stays Clean

An RX 12 Module Chain applies several modules in series and saves their order and settings as a reusable preset. Build it from defects you can hear, not from every module you own. Repair clipping and short impulses early, use hum/noise/reverb stages only when the source needs them, and place final level or loudness work last.

Validate the chain by rendering a working copy and listening to the result. Module Chain is an offline serial workflow: ordinary transport playback does not audition the stack, and switching a slot off does not undo an earlier render. Compare candidates from the same untouched source state at matched loudness. Files with different microphones, rooms, gain or damage need fresh settings checks before you reuse a preset.

What RX Module Chain actually saves

The iZotope Module Chain documentation defines it as serial processing: modules run from top to bottom with one Render command. You can start with a factory preset, add or remove slots, choose a module for each slot, reorder the signal flow, bypass individual stages and save the result as a custom chain preset.

The RX 12 edition comparison includes Module Chain and Batch Processor in Standard and Advanced. Elements has no standalone Audio Editor. Individual stages still depend on your edition: Leveler and Trim Silence require Advanced, while Loudness Control is available in Standard and Advanced. A preset cannot unlock a missing module.

Use the module search to find Module Chain, then use + to add a stage, its drag handle to reorder it, its power button to enable or disable processing, and View Module to edit that stage’s settings. Frequency Selection Settings can restrict a stage to a chosen frequency range; check those settings when recalling someone else’s preset. The manual explicitly excludes Wow & Flutter from Module Chain.

The current module-footer rules state that Preview, Compare and the other footer controls are unavailable when a module is loaded inside Module Chain. Tune a module outside the chain with its available audition controls, or render a short test copy. To hear the whole chain, render that copy and play it back. Do not mistake a power switch for a live A/B control on already processed audio.

A chain stores repeatable processing decisions; it does not store human judgment about every future file. A learned noise print, scene-dependent neural separation or threshold tuned to one narrator may be wrong for the next take. The Preview, Compare and History guide covers the evaluation controls that keep the preset honest.

There is no universal RX processing order

Order matters because every stage receives the output of the stage above it. A denoiser can react differently after clipping is repaired. Compression or loudness processing can raise mouth clicks and room noise. De-reverb may respond differently after broadband noise has been reduced.

Start with a provisional order: repair genuine waveform damage and isolated events, then evaluate persistent interference and room sound, and finish delivery level. The vendor’s 2018 repair-order guidance recommends impulse repair before hum or broadband reduction. Treat that as workflow guidance, not proof that every recording needs all five rows below. Compare an alternative order from the same source when the first result damages wanted sound.

Nine steps to build a safe Module Chain

1. Preserve an independent source. Keep an untouched audio-file copy and open a separate working copy. A duplicated region or playlist may still reference the same disk file.
2. Choose the test scope. Use the intended file tab with Composite View and Instant Process off. Include damaged and clean passages; exclude Trim Silence from a selection-only repair.
3. List confirmed problems. Add only available modules that solve audible defects. Check each slot’s frequency range and power state.
4. Tune stages outside the chain. Use an individual module’s Preview or Compare where available, or render a test copy. Save its approved settings as a module preset and load that preset in the matching chain slot.
5. Build the proposed order. Add each stage in signal order and check settings at the point where it receives audio. If testing has already processed the file, return to the untouched source state before applying the complete chain.
6. Render the first candidate. Confirm each stage is set for the intended repair, then use the chain’s Render control on a working copy. Play back the rendered result and check the complete recording or intended selection.
7. Test a stage without stacking renders. Keep candidate A, return to the same source state, disable one slot and render candidate B. Compare at matched loudness; disabling a slot alone cannot reverse processing already in the file.
8. Save a descriptive chain preset. Use Add Preset to save the validated order and settings. Include source, task, intensity and version; preserve a known preset before revising it.
9. Validate delivery and reuse. Export a new audio filename and reopen it. Check duration, alignment, channels, metadata and required loudness, then test a representative new file before any batch.

Useful preset names encode limits: lav_indoor_hum_light_v1 is safer than dialogue_master. The name reminds the next editor that a different microphone or exterior location needs a fresh check.

A conservative dialogue chain

For ordinary spoken word, start with only the failures present. De-clip comes first if peaks are actually flattened. Mouth De-click or De-click handles short impulses. De-hum follows when a tonal fundamental and harmonics persist. Voice De-noise or Dialogue Isolate addresses the remaining broadband or changing background. Level and loudness work come after repair.

De-reverb is optional. The RX 12 De-reverb manual allows either order with Spectral De-noise, so there is no vendor-mandated denoise-first rule. If one order sounds hollow or metallic, compare the other on a fresh copy and reduce the amount. Relearn the reverb profile from the audio actually reaching that stage, including direct sound, tails and room tone. The De-reverb tutorial explains those listening checks.

Do not automatically stack strong mouth, breath and sibilance processing. A voice needs articulation and air. Use the Mouth De-click guide to audition removed clicks with Listen during Preview in the individual RX 12 Editor module. Turn Listen off and approve normal speech before saving the module preset. The Listen footer is absent inside Module Chain; render a test copy to evaluate the complete chain. For plosives, the De-plosive manual specifically recommends processing before a high-pass filter.

A podcast or audiobook chain

Long-form speech rewards repeatability but punishes small artifacts repeated for hours. Group files by narrator, microphone and room. Build a light technical chain for each group, then check chapter starts, loud lines, whispers and edits before using it across the set.

In iZotope’s May 2026 podcast and audiobook example, Trim Silence precedes Dialogue Isolate, and those two stages are saved as a Module Chain. Leveler and Loudness Control finish the file only when no further creative editing is required. This example uses Advanced-only tools; Standard users can still build repair chains from their available modules.

Trim Silence ignores the current selection and always processes the entire audio file, according to the RX 12 Trim Silence manual. A short highlighted range is not protection. Keep it out of picture-locked dialogue, synchronized multitrack recordings and selection-only repairs. For independent podcasts or audiobooks, work on a copy and review every shortened pause for meaning, breaths and cut-off words. Threshold detects low-level material, not semantic mistakes or unwanted conversation.

Module Chain versus Repair Assistant

Repair Assistant analyzes the source and proposes processing. Module Chain stores a deliberate sequence and runs it again. Use Assistant to identify a plausible starting point, then inspect what it selected and whether each stage fits the defect.

The RX 12 Repair Assistant manual limits Open Module Chain to Advanced. Use Voice mode for spoken dialogue and Music mode for sung vocals or instruments. Opening the recommendations transfers a starting point; it does not make the new chain follow Assistant’s preview live. Avoid rendering Assistant and then applying its transferred chain again to the already treated file. Standard users can configure Module Chain manually.

A chain should not become a frozen Assistant result copied across unrelated recordings. Thresholds and reduction amounts that were mild on a loud studio voice can be destructive on quiet field dialogue. Reopen each slot, retune the representative file and remove modules that are not solving an audible problem.

The RX Repair Assistant guide separates analysis from approval. Its most useful contribution to Module Chain is a reasoned starting point, not permission to batch without listening.

How to prevent stacked artifacts

Each stage can sound acceptable alone while the series becomes hollow, gated or phasey. A gentle denoise removes a little ambience; De-reverb removes more tail; Dialogue Isolate separates again; final compression exposes the processing texture. Judge the sum, not the screenshots.

Make candidate A with the complete proposed chain. Keep it as a separate file or RX Document, return to the same source state, disable one slot and render candidate B. If B preserves natural speech without bringing back an audible problem, remove that stage. Repeat from the same source for the next comparison. Do not render B onto A: that would compare cumulative processing, and a disabled slot cannot restore what A removed.

Compare at matched loudness. Leveler, Normalize or Loudness Control can change level enough to bias the choice. Make repair-only candidates with finishing stages disabled before rendering, then evaluate the approved repair through its delivery chain. Check loudness and true peak against the actual destination’s specification; one podcast target is not a universal broadcast standard.

Use presets as baselines, not promises

Factory presets reveal intended combinations and parameter ranges. Custom presets save time on recurring sources. Neither knows whether today's file has a different noise floor, room, codec, microphone distance or speaker.

The common RX module controls allow presets to be added, renamed, imported and recalled. Keep a version number in the chain name and a short project note describing the test source. When a material change is justified, save a new version instead of silently overwriting the known one.

If recurring work needs several strengths, save separate light and more intensive variants with notes about their limits. A rescue preset should remain a deliberate choice for compromised material.

Module Chain versus Batch Processor

Module Chain applies a serial stack to the active file or selection. Batch Processor runs processing jobs across groups of files and adds destination, naming, output format, bit depth and dither decisions. They solve related but different workflow problems.

The RX 12 Batch Processor manual explains that its Module Chain area can recall a saved chain, while a Batch Processor preset can store both chain contents and output options. Choose the intended output folder and filename suffix explicitly. A chain preset alone is not the complete delivery configuration. Wow & Flutter is also excluded from Batch Processor.

Keep the original recordings separate from the output destination. Process one representative file, inspect its start/end, channel count, metadata and loudness, then try a small pilot group before the remainder. Grouping by microphone or narrator is only a starting point: listen for changing noise and room conditions within each file. Use the Batch Processor guide for input and output setup.

Selections and timing-sensitive modules

Check each module’s scope rather than assuming every slot follows the time selection. Ordinary repair stages can act on selected audio, and a slot can have its own frequency limits. Trim Silence is the important whole-file exception. Time-changing edits deserve a separate test and explicit downstream approval; resampling and bit-depth conversion also need the intended output specification.

For picture, games or multi-track music, preserve duration and alignment unless the downstream workflow expects a change. A clean line that returns a few frames short can break sync. Keep repair chains separate from editorial chains when timing must remain invariant.

A learned profile is source-specific. Tune a noise-only profile on noise, but train De-reverb with direct sound and tails; Assistant has its own content-based learning procedure. If an earlier stage changes the signal, revisit the profile used by the next stage. Recall the preset and render a test to confirm its settings rather than assuming it will learn every incoming clip automatically. Try adaptive processing where the chosen module supports it and compare across changing sections; adaptive does not mean artifact-free.

Save the edit, not only the preset

A preset remembers controls; an RX document can preserve the source, edits, history and view state. The RX file workflow documentation describes .rxdoc as an Editor-only archive that can become large when full-file processes accumulate.

Keep the untreated audio, chain preset and processed deliverable; retain an RX Document when you need editable history. The current Undo History documentation groups chain operations under a Module Chain parent. Return to the state before the complete chain when starting a new comparison; do not assume a single child undo removes every stage. History XML records edits and selections, but it is not the source audio or a substitute for an RX Document.

Do not confuse a saved preset with a reproducible master. Reproduction also requires the RX version, edition, source identity, sample rate, selection, external plug-in versions and output settings.

Save overwrites an opened WAV or AIFF on disk. Export a distinct filename for delivery and check whether you chose the whole file or Export Selection, then reopen it. Standard’s Editor handles mono/stereo; Advanced supports multichannel files up to ten channels per tab. Preserve required channel mapping and sync. Apply dither deliberately at the final integer bit-depth reduction. Avoid adding it both in a chain stage and again in export without a reason; inspect final peaks after conversion because noise shaping can raise them.

A music repair chain needs different assumptions

Do not copy a spoken-word chain onto a vocal, guitar or full mix merely because the meters look similar. Dialogue Isolate can change musical ambience, Mouth De-click can react to pick attacks and percussion, and aggressive De-reverb can shorten intentional space. Start with the performance problem: amp hum, fret squeak, headphone bleed, clipped peaks or one unwanted event.

For a vocal, test repair of obvious clipping and mouth clicks before tone shaping, then use only the noise or room stage the take requires. Guitar De-noise has separate Amp, Squeak and Pick sections. Amp targets static tonal hum or buzz; broadband hiss may need a denoiser. Keep useful pick attacks and fret sounds unless the performance calls for their reduction. For a stereo mix, preserve image and transients and avoid a default dialogue chain.

Music presets should name the source and musical role, not only the defect. A chain approved for an exposed acoustic-guitar intro may be too aggressive once drums enter, while a backing-vocal preset can damage a lead. Test the cleanest and busiest passages, then leave local spectral events for selections outside the global chain.

Final Module Chain quality gate

• Every active module solves a confirmed audible problem.
• The order was compared on a representative worst-case passage.
• Structural and impulse repair precede broad reduction where appropriate.
• Level, loudness and dither decisions remain at the delivery end.
• Each slot was tested with a fresh render from the same source state, and retained only when useful.
• The complete chain was compared at matched loudness on clean and damaged passages.
• The preset name records source, task, intensity and version.
• A new file is tested manually before any batch or production reuse.

The Learn iZotope RX hub connects Module Chain to selections, spectrogram reading, Repair Assistant and the Editor fundamentals that make automation safe.

Frequently asked questions

What does RX Module Chain do?

It renders available RX modules in a saved serial order. You can add, remove, reorder, disable and configure stages, then save the chain as a preset. Check each module’s scope: Trim Silence ignores the selection and processes the whole file.

What is the best RX processing order?

There is no universal order. A safe baseline repairs waveform damage and impulses first, persistent noise next and delivery level last, but the source and an A/B comparison decide.

Should De-noise come before De-reverb?

The RX 12 De-reverb manual permits either order with Spectral De-noise. If both are needed, compare from the same source state and retune the profile for the signal reaching each stage. Keep the less damaging result.

Can I save a Module Chain preset?

Yes. Save a custom preset after every stage is validated. Use a descriptive source/task/intensity name and create a new version when settings materially change.

Can Module Chain process multiple files?

Use Batch Processor for groups of files. Validate the processing order on one representative file, then use a new destination and pilot batch before the full set.

Is Module Chain the same as Repair Assistant?

No. Repair Assistant analyzes audio and proposes settings; Module Chain stores and renders a sequence. Advanced can transfer Assistant suggestions with Open Module Chain. Standard users can build chains manually. A transferred chain still needs its own rendered test.

Why does my chain sound metallic or hollow?

Denoise, separation and de-reverb can accumulate artifacts. Keep one rendered candidate, return to the same source state, disable a slot and render another. Compare at matched loudness. Switching a slot off does not undo an earlier render.

Should Trim Silence be in every podcast chain?

No. This Advanced module ignores the selection and processes the whole file, shortening low-level regions. Exclude it from picture-locked or synchronized material, and review pauses and word endings on an independent output copy.