What Are Peak Levels and Why They Matter in Dialogue

In audio production, peak levels represent the highest instantaneous amplitude of a signal. When recording dialogue, these peaks come from sudden vocal emphasis, plosives, sibilance, or even mechanical noise like mic bumps. If peaks exceed the system’s headroom, they cause clipping—a hard, distorted sound that cannot be corrected in post. Even when clipping is avoided, excessively high peaks force an engineer to reduce overall gain, pushing quieter passages closer to the noise floor. The goal of managing peaks is to reduce the crest factor (the difference between average and peak levels) without making speech sound squashed or lifeless. This preserves intelligibility: the listener understands every word without strain, even in noisy environments or on consumer devices like smartphones and laptops.

Peak levels are measured in dBFS (decibels relative to full scale) in digital systems, with 0 dBFS representing the maximum possible level before clipping. In practice, you want to keep peaks well below 0 dBFS, typically around –6 dBFS or lower during recording, to leave headroom for processing. Understanding peak levels is the first step toward controlling them effectively.

Understanding Dynamic Range and Its Role in Speech Clarity

Dialogue naturally exhibits a wide dynamic range. A speaker might whisper, then shout, or emphasize a word with sudden force. This variation conveys emotion and meaning. However, in recorded media—whether for podcasts, films, or voiceovers—extreme dynamic range causes problems. Quiet parts become inaudible, and loud peaks cause listener fatigue or equipment overload. The goal is not to eliminate dynamics but to control them. The techniques that follow reduce peak levels while preserving the natural ebb and flow of speech, ensuring the final mix sounds both polished and human.

Speech intelligibility depends on preserving the transient information in consonants—especially plosives and fricatives—while taming the energy that causes distortion. A well-controlled dynamic range keeps dialogue clear at any listening level. The human ear is remarkably good at filling in missing information, but only up to a point. If peaks are crushed or muddled, comprehension suffers. This is why transparent peak control is essential for professional dialogue production.

Technique 1: Compression – The Foundation of Peak Control

Compression reduces the level of audio that exceeds a set threshold by a certain ratio. For dialogue, a gentle ratio (2:1 or 3:1) and a well-chosen threshold (around –10 to –20 dBFS) tame the loudest syllables without killing transients. The attack setting is critical: too fast (0.5 ms) can crush the beginning of words and dull the voice; too slow (20 ms) lets peaks through. A medium attack (5–10 ms) works well for most dialogue. Release should be fast enough to recover between phrases (50–100 ms), avoiding pumping or breathing artifacts. Use a compressor with a soft knee for smoother gain reduction. Always check that compression is not introducing noise or making dialogue sound distant.

For example, on a voiceover recording with peaks at –8 dBFS and an average level around –20 dBFS, a threshold of –14 dBFS with a 3:1 ratio will reduce the loudest sections by about 2–3 dB, bringing the crest factor down without noticeable artifacts. The makeup gain restores the overall level, resulting in a more consistent track. Over time, your ears learn to hear when compression is too aggressive: the voice loses its forward energy and sounds like it is coming from inside a box.

Parallel Compression for More Natural Dynamics

Also known as New York compression, this technique blends a heavily compressed version of the signal with the dry original. To set it up, create a bus with heavy compression (4:1 or higher, with fast attack and release) and blend it back into the dry signal at around 20–40% wet. This adds body and consistency without making dialogue sound over-processed. Start with a 50/50 mix, then adjust to taste. Parallel compression is especially effective for dialogue that needs to cut through a dense mix, such as in film or television.

Serial Compression for Tighter Control

Using two compressors in series, each doing a small amount of work (2–3 dB of gain reduction), often sounds more natural than one compressor doing 6 dB of reduction. The first compressor catches the broad peaks, while the second handles the remaining fine details. This approach mimics the way analog consoles were often set up and remains a staple in professional dialogue chains.

Technique 2: Limiting – The Safety Net for Peaks

A limiter is a compressor with a very high ratio (10:1 or higher) and a fast attack. It acts as a brick wall: no signal can exceed the ceiling level. Use a limiter on the dialogue bus to catch any remaining peaks after compression. Set the ceiling to –1 dBFS (for digital delivery) to avoid intersample peaks. True peak limiting is even better because it accounts for the actual analog waveform reconstructed from digital samples. Limiters are safety devices, not creative tools. Over-limiting causes audible distortion and pumping, so use only a few dB of gain reduction at most.

For loudness standards (like –23 LUFS for broadcast or –16 LUFS for podcast), a limiter at the final stage ensures compliance while protecting against overs. The EBU R128 standard recommends a maximum true peak of –1 dBFS for broadcast, making true peak limiting a requirement for professional delivery. Always check your limiter's output with a true peak meter to verify that no overs are slipping through.

Technique 3: Manual Gain Riding – Old-School Precision

Before compressors became ubiquitous, engineers used faders to ride levels in real time. This method remains one of the most transparent ways to reduce peaks. Gain riding (or gain automation) involves lowering the volume during loud phrases and raising it during quiet ones. In a DAW, you can draw volume automation curves frame by frame. For dialogue, this is especially useful for fixing variations in microphone distance or vocal intensity. The advantage: no artifacts, no tonal change. The disadvantage: it is time-consuming. But for critical projects (feature films, high-end podcasts), nothing beats hand-drawn automation.

A practical approach is to first do a pass where you normalize each phrase so that the average level stays consistent, then apply light compression to smooth any remaining peaks. Work in a quiet room with good monitors, and listen at a moderate level so you do not over-correct. Use clip gain (or gain envelopes) rather than fader automation when possible, as this preserves the fader for final mixing moves. Many professional dialogue editors spend as much time on gain riding as they do on any other processing step.

Technique 4: Equalization (EQ) to Reduce Peaks Before They Happen

Not all peaks are created equal. Many problematic peaks come from specific frequency ranges, and EQ can address them before they reach the compressor:

  • Plosives (p, b, t) concentrate energy below 150 Hz. A high-pass filter at 80–100 Hz reduces those low rumble peaks without affecting vocal clarity. For voices with excessive low-end energy, set the filter as high as 120 Hz, but check that the voice does not sound thin.
  • Sibilance (s, sh, ch) lives around 5–10 kHz. A gentle de-esser or a narrow cut in that range tames harsh peaks.
  • Nasal resonances in the 500–800 Hz area add boxiness that can trigger compressors prematurely. A slight cut (2–3 dB with a narrow Q) smooths the signal before it hits compression.
  • Presence peaks in the 2–4 kHz range can make dialogue sound aggressive. A subtle dip here reduces listener fatigue while keeping intelligibility high.

Apply EQ before compression so the compressor reacts to a more balanced signal. This prevents it from over-reacting to boomy or piercing sounds. For example, a voice with a strong 120 Hz resonance will cause the compressor to gain-reduce on every plosive, dulling the entire performance. By cutting that resonance first, the compressor works more evenly. For a deeper dive into EQ for vocals, see iZotope’s EQ guide for dialogue.

Technique 5: De-essing – A Specialized Peak Tamer

Sibilant peaks are often the loudest and most distracting elements in a dialogue track. A dedicated de-esser targets only the sibilant range, reducing its level while leaving the rest of the voice untouched. Most de-essers work as frequency-selective compressors. Set the frequency to the sibilance center (usually 5–8 kHz for men, 6–9 kHz for women) and adjust the threshold so only the harsh esses trigger reduction. Use a split-band de-esser for better artifact control; these devices split the signal into two frequency bands and compress only the sibilant band, leaving everything else unprocessed.

For extreme sibilance, you can use two de-essers in series: one set to catch the main sibilant range and another set to a narrower band for the worst offenders. Alternatively, use volume automation on individual esses if the de-esser is causing audible pumping. Many engineers prefer de-essing before compression because the compressor can then work more evenly across the signal. However, if the compressor itself creates sibilant artifacts, try de-essing after compression instead. Experiment to find what works for each voice.

Technique 6: Multiband Compression – Precision Peak Control Across Frequencies

Standard compression treats the entire frequency range equally, but a multiband compressor splits the signal into bands (low, mid, high) and compresses each independently. This is powerful for reducing peaks in one area without squashing another. For example, if a speaker’s low frequencies are overly energetic, compress only the low band (20–200 Hz) with a moderate ratio (2:1) and a fast attack. Meanwhile, the mid and high bands remain untouched, preserving intelligibility and air.

Multiband compression also helps reduce muddiness or honkiness in specific frequency ranges. A voice that sounds boxy in the 300–500 Hz range can be tightened by compressing that band by 2–3 dB. Use multiband compression sparingly—over-processing makes dialogue sound splintered and unnatural. A good rule of thumb is to start with the crossover points at 200 Hz and 4 kHz, and apply no more than 3 dB of gain reduction in any band. For more details, see Avid’s tips on multiband compression.

Technique 7: Microphone Technique and Recording Discipline

The best way to reduce peak levels is to prevent them at the source. Educate voice actors or interview subjects to maintain a consistent distance from the microphone (typically 6–12 inches). Use a pop filter to soften plosives. Position the mic slightly off-axis to reduce sibilance and room reflections. Choose a dynamic microphone (such as the Shure SM7B or Electro-Voice RE20) for podcasting and voiceover—these handle high SPL better and have less sensitivity to off-axis noise, which helps control peaks. Condenser mics capture more detail but also more transient energy; they require more careful gain staging and often benefit from a pad switch to prevent overload.

During recording, set preamp gain so the loudest passages peak no higher than –6 dBFS (digital) or 0 VU (analog). This leaves headroom for post-processing. Use a hardware compressor with a gentle ratio during tracking if you are confident in your setup, but many engineers prefer to capture clean audio and compress later. Always monitor the recording with peak meters and listen for distortion. A good recording session reduces the amount of compression and limiting needed later, resulting in a more natural final product.

For remote interviews or location recording, educate your talent on mic technique. A simple script with reminders about distance and volume can dramatically improve the quality of the source material. Remind them to avoid moving their head during loud phrases and to keep the microphone at a consistent angle. These small habits pay dividends in post-production.

Practical Workflow: Combining Techniques for Maximum Effect

Here is a step-by-step workflow that integrates the above methods into a coherent process:

  1. Gain stage first: Set input levels to peak around –12 dBFS average, –6 dBFS on the loudest lines. This ensures clean, unclipped audio with enough headroom for processing.
  2. Edit out breaths and clicks that can trigger compressors unnecessarily. Use spectral editing or manual deletion for the best results.
  3. Apply a high-pass filter at 80 Hz (or higher for voices with excessive low-end energy). This removes rumble and plosive energy before compression.
  4. Use gentle compression with a 2:1 ratio and a threshold that yields 3–6 dB of gain reduction on average. Attack around 5–10 ms, release around 50–80 ms, with a soft knee.
  5. Add a de-esser set to the sibilant range (5–8 kHz for men, 6–9 kHz for women). Adjust threshold so only the harshest esses trigger reduction.
  6. Draw volume automation to smooth out remaining inconsistencies. Focus on phrases that are significantly louder or quieter than the average.
  7. Apply a final limiter set to –1 dBFS true peak (or –2 dBFS for delivery specifications like broadcast). Use only 1–3 dB of gain reduction.
  8. Check loudness with a meter (e.g., –16 LUFS for podcasts, –23 LUFS for TV or film). Adjust makeup gain or compression to hit the target.

Each step should be applied subtly. If you find yourself adding more than 6 dB of gain reduction in any single stage, revisit earlier steps. The goal is cumulative, minimal processing. Listen on multiple playback systems—headphones, laptop speakers, and car audio—to verify that the dialogue remains clear and natural.

Common Pitfalls to Avoid

  • Over-compression: Squeezing the life out of dialogue makes it sound dull, flat, and fatiguing. The voice loses its natural variation and emotional range. If you notice the waveform looking like a solid block, you have gone too far.
  • Too much limiting: More than 2–3 dB of limiting starts to produce audible distortion and pumping. Use limiting as a safety net, not as a primary level control tool.
  • Ignoring the listening environment: If you mix on headphones or untreated speakers, you may misjudge peaks and dynamics. Check on multiple playback systems and at different volumes.
  • Not checking for phase issues: When processing stereo or multitrack recordings, ensure that EQ and multiband compression do not introduce phase shifts that smear the sound. Use linear-phase EQ if necessary.
  • Neglecting the noise floor: Reducing peaks is pointless if the noise floor rises audibly. Use noise gates or expanders carefully, and avoid excessive makeup gain that amplifies background noise.
  • Processing in the wrong order: EQ before compression, de-essing before or after compression depending on the voice, and limiting last. Changing the order can produce very different results.

Why Intelligibility Matters Beyond Peak Levels

Managing peak levels is critical, but it works hand-in-hand with other elements of audio quality. A well-processed dialogue track should sound natural even at low volumes. The listener should not notice the processing; they should just understand every word. This means preserving the transients of consonants (which carry information) while taming the loud ones. Use transient shaping tools if needed, but often simple gain riding plus gentle compression achieves the best result. Remember that dialogue is for storytelling—do not let technical fixes get in the way of emotion.

Intelligibility is especially important in noisy environments, such as while driving or in a crowded room. A dialogue track with well-controlled peaks and a consistent level will cut through background noise better than one with wide dynamic swings. The ANSI S3.5-1997 standard for speech intelligibility emphasizes the importance of preserving consonant energy, which is easily masked by peaks in the low frequencies. By reducing low-frequency peaks with EQ and compression, you preserve the clarity of consonants and improve overall comprehension.

For further reading on loudness standards and speech intelligibility, see the EBU’s loudness guidelines and the AES recommended practice for dialogue intelligibility. These resources provide technical background on metering, loudness targets, and best practices for broadcast and film delivery.

Final Thoughts

Reducing peak levels in dialogue recordings without sacrificing intelligibility is a balancing act. It requires understanding the nature of speech, using the right tools in the right order, and trusting your ears. Start with recording discipline, then apply EQ, compression, and limiting in moderation. Use gain riding for finesse, and always monitor in context. With practice, you will achieve clean, natural-sounding dialogue that holds the audience’s attention from first word to last.

The techniques described here are not one-size-fits-all. Every voice is different, every recording environment is unique, and every delivery format has its own requirements. Learn to listen critically, experiment with settings, and develop a workflow that works for you. The goal is dialogue that sounds effortless—clear, present, and emotionally engaging, without any hint of the processing that made it possible.