Understanding the Role of Noise Gates in Dialogue Production

In audio post-production, background noise is one of the most persistent challenges—whether it's the hum of an air conditioner, the rustle of clothing, or the low rumble of traffic seeping into a vocal track. A noise gate is a specialized processor designed to automatically reduce or silence audio below a set level, effectively "closing the gate" on unwanted noise when the desired signal (like speech) is absent. When used properly, a noise gate can transform a hissy, roomy dialogue track into a clean, focused recording without stripping away the natural texture of the voice.

However, a poorly configured gate can be just as damaging as the noise itself—introducing clicks, unnatural fade-ins, or robotic articulation. The key lies in understanding the four core parameters and how they interact with the unique dynamics of human speech. This article walks through every setting, offers production-ready tips, and explains how to combine gating with complementary tools for transparent noise reduction.

Core Parameters: Threshold, Attack, Release, and Hold

Before diving into workflow, it's essential to grasp how each noise gate control behaves in dialogue contexts. While the same fundamentals apply across music and post, speech presents specific challenges because of its wide dynamic range—from soft consonants to loud plosives—and the natural pauses that occur between words and phrases.

Setting the Threshold

The threshold determines the level at which the gate opens (passes audio) or closes (attenuates or mutes). Set it too low, and the gate stays open almost all the time, defeating the purpose. Set it too high, and soft speech endings (like the trailing "s" in "misses") get chopped off, creating an unnatural, staccato effect. The ideal threshold sits just above the noise floor—usually 2 to 6 dB higher than the ambient noise level—but below the quietest part of the speech you want to keep. A good technique is to play the quietest syllable from your dialogue, watch the level meter, and set the threshold about 3 dB below that point. This ensures that even whispered phrases or breathy consonants trigger the gate reliably.

Attack Time: Preserving Speech Onsets

Attack controls how quickly the gate opens once the signal crosses the threshold. For dialogue, a fast attack (0.5 to 2 ms) is generally desirable because plosives (like "p" or "b") have a very fast transient. If attack is too slow, the first few milliseconds of each word can get dulled or missed entirely, making speech sound muffled or distant. However, setting it too fast (under 0.1 ms) on a track with a high noise floor can cause the gate to open on very short bursts of noise—like a sudden rustle—introducing ugly clicks. A safe starting point for clean dialogue is 1-3 ms; adjust downward only if the initial transients still feel soft.

Release Time: Avoiding Choppy Endings

Release determines how long after the speech falls below the threshold the gate takes to close. This is arguably the most critical parameter for naturalness. A release that is too fast (under 10 ms) will cause each word to end abruptly, cutting off the natural decay of sibilants or the tail of a breath. A release that is too slow (over 200 ms) may leave the gate open during short pauses between words, allowing noise to pass through and making the track feel less clean. For most dialogue, a release between 40 and 80 ms works well. Tune it by listening to the end of a sentence: you want the gate to close smoothly, as if the noise naturally fades by the time the next word arrives. If you hear a "chop" at the end of phrases, lengthen the release. If you hear noise bleeding into the gap between sentences, shorten it.

Hold Time: Preventing Chatter During Pauses

Hold keeps the gate fully open for a brief period after the signal drops below the threshold, before the release phase begins. This is extremely useful in dialogue because it prevents the gate from closing during the tiny pauses between words in quick speech. For example, in the phrase "I think I can," the natural gap between "think" and "I" is only about 50-100 ms. Without hold, the gate could close and reopen, causing a "chattering" effect. Adding a hold of 20-50 ms locks the gate open during those micro-pauses, letting the release handle the longer silence at the end of a sentence. Hold is also valuable when you want to preserve the natural rhythm of speech without making the gate sound too busy.

Practical Workflow: Calibrating the Gate on a Dialogue Track

Step-by-step method for setting up a noise gate on a dialogue clip (for example, from a podcast or narration recorded in a slightly noisy room):

  1. Listen to the noise floor soloed. First, identify the loudest consistent background noise (AC hum, traffic, fan, etc.). Note its peak level on your DAW's meter.
  2. Set threshold slightly above that noise floor. Typically 3-6 dB higher. Confirm that the gate stays closed during silent sections but opens consistently when speech begins.
  3. Attack at 1-3 ms. If the speech sounds dull at the very start of words, reduce attack to 0.5 ms. If you hear clicks from short noise bursts, increase attack to 3-5 ms.
  4. Hold at 20-50 ms. Start with 30 ms. Listen to a phrase with short pauses—if you hear the gate opening and closing between fast syllables, increase hold until the chatter disappears.
  5. Release at 40-80 ms. Listen to the end of a sentence. If the tail is chopped, increase release until the noise fades smoothly. If the next word is preceded by too much noise bleed, decrease release.
  6. Fine-tune with the dialogue playing. Bypass and engage the gate repeatedly to compare. The gated version should sound cleaner but not noticeably processed.

If the track still has audible noise during the speech itself—that is, the noise is louder than the gate's threshold—the gate alone cannot help. That noise is "in-band" with the dialogue and must be reduced with EQ, spectral editing, or a noise reduction plugin before or after gating.

Combining the Gate with Compression, EQ, and Noise Reduction

A noise gate is only one piece of a complete noise control strategy. Relying solely on gating often leaves residual noise that sits at the same level as quieter syllables. Here is how other tools work with the gate to achieve cleaner results while preserving naturalness:

Compression Before the Gate

If your dialogue has a wide dynamic range (e.g., a soft-spoken guest with loud outbursts), a compressor placed before the gate can even out the level. This makes the threshold easier to set because the quietest parts are lifted closer to the average level. A gentle 2:1 or 3:1 ratio with medium attack and release will smooth out peaks without making the voice sound unnatural. Then the gate can work with a lower threshold that still catches background noise during pauses.

EQ to Reduce Noise Before Gating

Applying a high-pass filter (around 80-120 Hz) removes rumbles and low-frequency noise like air conditioner drones. A gentle low-pass filter (around 8-10 kHz) can reduce hiss or high-frequency hum. These EQ moves reduce the noise floor itself, making the gate's job easier and allowing a higher threshold without cutting into speech frequencies. Be careful not to over-EQ—dialogue can sound thin if you remove too much low end or too much air.

Dedicated Noise Reduction as a First Stage

For persistent broadband noise (like fan noise or crowd rumble), a noise reduction plugin (e.g., iZotope RX De-noise or Waves WLM) can be used to sample and reduce the noise floor. Then the gate is applied to handle any residual or transient noise that wasn't removed. This two-stage approach gives you the cleanest result without the artifacts that pure noise reduction can cause when pushed too far.

Common Mistakes and How to Avoid Them

Many audio engineers—especially those new to dialogue processing—make these errors when using noise gates:

  • Threshold too high: Cutting off the natural tails of words, making speech sound clipped. Solution: lower the threshold until all syllables trigger the gate reliably.
  • Attack too fast: Causing clicks or "spitty" sounds on plosives. Solution: increase attack slightly (to 2-4 ms) to let the transient pass before the gate opens—this reduces harshness.
  • Release too fast: Creating a choppy, robotic rhythm. Solution: increase release until the gate's closure feels smooth and natural.
  • Release too slow: Allowing noise to bleed through between sentences. Solution: decrease release gradually until noise is suppressed but speech still feels fluid.
  • No hold time: Causing the gate to chatter on fast speech. Solution: add hold (20-50 ms) to bridge micro-pauses.
  • Gating only: Ignoring the fact that noise occurring during speech cannot be gated out. Solution: use EQ, compression, or noise reduction to lower the noise floor before gating.

Advanced Techniques: Sidechain Gating and Keying

In many production scenarios—especially in film or broadcasting—you may want to clean up one track based on the level of another. This is called sidechain gating. For example, if you have a boom microphone and a lavalier, you can sidechain the boom to the lav: the boom's gate opens only when the lav's signal exceeds the threshold. This ensures that only the lav (which captures clearer dialogue) is heard when the boom picks up too much room noise. Sidechain gating is also used for de-essing (gate triggered by high frequencies) or for dynamic EQ applications. Most DAWs and hardware processors allow you to route a key input; the gate listens to the key source but processes the audio source. This is an advanced but powerful way to preserve naturalness because the gate's timing is tied to the cleanest source of speech.

Real-World Scenarios: Podcast, Voiceover, and Field Dialogue

The optimal gate settings vary depending on the recording environment and the performance style. Here are three common situations with recommended starting points:

Podcast (Close-Mic, Controlled Room)

Typically low noise floor, but may have pops and chair squeaks. Use a fast attack (1 ms), moderate hold (30 ms), and release around 50 ms. Threshold should be set low (e.g., -40 dB to -50 dB) because the noise floor is low. The gate here is more about cleaning up mouth clicks and page rustle than suppressing broadband noise.

Voiceover (Studio with High-Quality Mic)

Very clean source, but the voice may have dynamic variation. Use a threshold just above the noise floor (around -55 dB to -60 dB), attack 2 ms, hold 20 ms, release 60 ms. The gate should be almost transparent; its main job is to prevent any silent hiss from building up during long pauses.

Field Dialogue (Boom Mic, Untreated Room)

Higher noise floor (traffic, wind, room tone). Use a higher threshold (e.g., -30 dB to -40 dB) but compensate by combining with a noise reduction plugin first. Use a slower attack (3-5 ms) to avoid opening on wind gusts, and a faster release (40 ms) to cut off noise quickly when speech stops. Hold at 40 ms to allow for quick sentence breaks. This is the hardest scenario—often you'll need to automate the threshold or use a multi-band gate to target specific frequencies.

When Not to Use a Gate

Not every dialogue track benefits from gating. If the background noise is constant and relatively quiet (e.g., a soft room tone), a gate can create an unnatural "pumping" effect as the room tone disappears during pauses and reappears during speech. In such cases, noise reduction or a high-pass filter is a better solution. Also, if the speaker's voice is very breathy or has a lot of sibilance, the gate may open and close on those sounds, causing distraction. Consider using a de-esser or equalizing the sibilance before gating. Finally, if the dialogue is heavily processed with compression (e.g., FM radio-style), the gate may cause obvious level changes because the compression has already reduced the dynamic range. In that case, gating should be applied before compression.

Conclusion: Balancing Cleanliness and Character

When applied with care, a noise gate can clean up dialogue without making it sound processed or unnatural. The secret is to treat the gate as a precision instrument—fine-tuning threshold, attack, release, and hold to match the natural rhythm of speech. Always listen in context: gating a soloed track may sound fine, but in a mix, the gate's actions can become audible. Use sidechain, EQ, and compression as complementary tools. And remember that no gate can fix a poor recording. Good source capture remains the foundation; the gate is there to polish, not to salvage.

For further reading, explore the detailed guides on Sound on Sound and the Whole Hog Audio blog, which provide additional examples and technical depth. With practice, you'll develop an intuition for when to gate, how much, and where to draw the line between a clean track and a sterile one.