Quick answer: for steady hiss, fan noise or HVAC wash, select a passage that contains only the unwanted sound, open Tools → Noise Reduction, capture the noiseprint, then compare Preview and Bypass with conservative settings. Stop when the noise no longer distracts. If the voice turns metallic, watery, dull or starts pumping, you have already gone too far.
This is part of the Sound Forge audio-restoration path. The job here is broadband background noise. Clicks, clipping, mains hum and background music are different defects; treating all of them with one denoiser is how clean recordings become thin and artificial.
Identify the Noise Before You Process It
The current Sound Forge 2026 product page lists broadband noise reduction, DeHisser, DeClicker/DeCrackler and DeClipper. Those tools overlap, but they do not solve the same problem.
- Steady hiss, fan or HVAC noise: start with broadband Noise Reduction and a clean noise-only sample.
- Tape or preamp hiss: compare Noise Reduction with DeHisser; keep whichever preserves consonants, cymbals and room tone better.
- Clicks and crackle: use Click and Crackle Removal or DeClicker/DeCrackler, not broadband reduction.
- Clipped peaks: use Clipped Peak Restoration or DeClipper. A noiseprint cannot reconstruct a flattened peak.
- A fixed electrical tone: inspect it in Spectrum Analysis before applying a narrow filter. Don’t notch a memorized frequency list without measuring the actual file.
- Background music or another voice: that is source separation, not steady-noise removal. A classic noiseprint cannot reliably isolate overlapping programme material.
A noise gate is not a substitute for broadband reduction. A gate can turn low-level noise down during pauses, but the noise remains under speech or music and may become more obvious as it switches in and out. Use the noise-gate workflow only when that behaviour fits the recording.
Where Noise Reduction Lives in Sound Forge 2026
For the classic plug-in, open Tools → Noise Reduction or use the Noise Reduction button on the Tools toolbar. The current Sound Forge 2026 Tools toolbar help separately documents Noise Reduction, Click and Crackle Removal, Clipped Peak Restoration and Audio Restoration.
The newer Sequoia restoration effects — DeHisser, DeClicker/DeCrackler and DeClipper — live under Effects. The current Repairing Audio help also mentions iZotope RX plug-ins and SpectraLayers. There is an important purchasing caveat: the live Sound Forge sales matrix does not name a specific RX module or SpectraLayers entitlement. Verify the license and installed modules you actually have before following an RX-specific tutorial.
That distinction matters because old search results often say that RX Elements or a particular De-noise module is bundled with every copy. Some of those pages describe a MAGIX-era package. In Sound Forge 2026, use the installed plug-in’s own ? help for its controls, and treat optional plug-ins as optional until your account or installer proves otherwise.
Classic NR-2.0 Workflow: Capture a Clean Noiseprint
The noiseprint is a frequency profile of the unwanted sound. It works best when the problem is reasonably steady and the sample contains no wanted audio. A longer selection is not automatically better: the useful selection is the cleanest representative passage you can find.
- Keep an untouched source. Work on a copy or save a new version before destructive processing.
- Select noise only. Use room tone, fan wash or hiss with no word ending, breath, music or transient inside the selection.
- Open Tools → Noise Reduction. In the classic dialog, switch to the Noiseprint controls if needed.
- Enable Capture noiseprint and preview the selected region. The classic plug-in analyses that selection to build the profile.
- Extend the target. The long-standing NR-2.0 workflow uses the context command Select All Data after the profile is captured. If your installed build behaves differently, use its ? help instead of forcing an old screenshot’s sequence.
- Preview a stress-test passage. Include wanted audio, a pause, a quiet consonant or note ending and a louder section.
- Compare Preview and Bypass. Apply only when the wanted signal remains stable.
NR-2.0 Modes: Preview the File, Not a Recipe
Legacy NR-2.0 documentation describes four reduction types, Mode 0 through Mode 3. Mode 0 removes the most noise but carries the highest artifact risk, Mode 2 is the legacy default, and Mode 3 is generally the gentlest. The best result still depends on the source, so audition all available modes rather than treating one as the voice, music or vinyl preset.
There is no universal best reduction amount. The right amount is the lowest setting that makes the noise stop drawing attention without changing the wanted sound. A number copied from a clean voice-over can damage a cassette transfer; a setting that works on steady fan wash can pump when the air conditioner cycles.
Use Keep residual output when the classic dialog provides it. This lets you hear what the process is removing. Noise is expected. Recognisable syllables, breaths, musical notes or ambience mean the profile or reduction is taking wanted content too. Back off, capture a cleaner sample or split the file into sections.
Choose the Repair That Matches the Defect
Room tone, HVAC and fan wash: capture a clean profile and use one conservative broadband pass. If the noise changes between sections, process those regions separately instead of making one profile fight the whole file.
Tape hiss: compare Noise Reduction with DeHisser. Listen to sibilants, cymbals, string texture and room ambience. Leaving a little natural hiss is better than replacing it with metallic “musical noise.”
Mains hum: use Spectrum Analysis to identify the actual fundamental and harmonics. Then use the narrowest correction that solves the audible problem. The power system may suggest where to look, but it does not tell you which peaks are harmful in this recording.
Clicks and crackles: use the dedicated click tools. For an isolated glitch, a local repair is safer than processing an entire programme. The click-and-pop guide covers selection, automatic repair and manual verification.
What to Do When There Is No Clean Noise-Only Passage
Do not teach NR-2.0 a profile contaminated with speech or music and hope the mode setting will fix it. First look at the head, tail and pauses between phrases. If every quiet area contains wanted sound, use smaller regions, an adaptive denoiser you actually own, or a spectral editor appropriate to the defect.
The official 2026 repair help tells users to open each optional plug-in’s ? help. Follow that advice. “Learn,” “Adaptive,” “Voice” and similar controls do not mean the same thing across products or versions, and the current purchase page does not promise one universal RX interface.
For a new recording, capture room tone before people move, the HVAC changes state or the microphone position shifts. That gives restoration a representative profile later. The Sound Forge recording guide covers the input and monitoring checks that prevent avoidable noise at source.
Artifact Check: Know When to Stop
Metallic or watery voice: the process is removing harmonics that belong to the speaker. Lower the reduction, try a gentler mode or capture a cleaner profile.
Dull music or softened consonants: use residual output and listen for wanted high-frequency detail. If you can recognise it in the removed signal, back off.
Pumping or breathing: the noise is changing, the profile is contaminated or the reduction is reacting too visibly between phrases. Split the file into sections or use an appropriate adaptive tool.
A strange silent hole between phrases: check whether a gate is also active. Broadband reduction should lower a noise floor; a gate can make it disappear abruptly and then return under the next word.
The finish line is not digital silence. It is a recording in which the noise stops competing with the content. A little stable room tone is often less distracting than aggressive processing that announces itself on every breath.
Cleanup Order: Repair Before Final Loudness
Keep the original, repair clicks or clipping where necessary, reduce broadband noise, then set final peak or loudness targets. Normalizing first raises the wanted signal and the noise together; it does not improve their ratio. Do the restoration decision on the cleanest source you have, then verify the rendered file and export the delivery copy.
If several processes are needed, preview the entire chain before rendering. A denoiser, EQ and limiter can each sound acceptable alone but become too aggressive together. Save a new working version before every destructive step so the only recovery strategy is not Undo.
Frequently Asked Questions
Where is Noise Reduction in Sound Forge 2026?
Open Tools → Noise Reduction or use the Noise Reduction button on the Tools toolbar. The current help places Sequoia DeHisser, DeClicker/DeCrackler and DeClipper under Effects, while optional licensed plug-ins appear through the FX Plug-In menu.
How do I capture a noiseprint in Sound Forge?
Select a representative passage containing only the unwanted sound, open Tools → Noise Reduction, enable Capture noiseprint and preview the selection. After capture, extend the target to the intended region or use the classic Select All Data workflow, then compare Preview and Bypass before applying.
What is the best Sound Forge noise-reduction setting?
There is no universal setting. Use the lowest reduction that makes the noise stop distracting without changing the wanted signal. Audition each available mode, compare Bypass and listen to residual output when available.
Why does noise reduction make a voice sound underwater?
The profile or reduction is removing wanted harmonic content with the noise. Lower the amount, try a gentler mode, capture a cleaner noise-only sample or process changing sections separately.
Should I remove background noise before normalizing?
Usually yes. Keep the untouched source, repair and reduce noise first, then set final peak or loudness. Normalizing raises the wanted audio and noise together; it does not improve the signal-to-noise ratio.
Can Sound Forge remove background music from speech?
Classic broadband noise reduction cannot reliably separate changing music from overlapping speech. That job needs source separation or spectral editing, and the result depends on the material and the specific tool you own.
Does Sound Forge 2026 include iZotope RX?
The current Sound Forge repair help mentions included RX plug-ins, but the live sales matrix does not name a specific RX module or edition. Verify your Boris FX account, license and installed plug-ins before relying on an RX workflow or an older bundle description.
What should I use for clicks, hum or clipping?
Use Click and Crackle Removal or DeClicker/DeCrackler for clicks, Spectrum Analysis before a narrow hum correction, and Clipped Peak Restoration or DeClipper for clipping. Those are different defects from steady broadband background noise.
For tonal noise, continue with the hiss-and-hum diagnosis guide; for exact frequency confirmation, use the Spectrum Analysis workflow.
Last fact-checked: August 2, 2026 against the current Sound Forge 2026 product, pricing, Tools toolbar, Repairing Audio and processing-dialog documentation.