Quick answer: start Fairlight dialogue EQ flat, name the audible problem, make one move, and compare at matched loudness. Use clip EQ for one recording that differs, track EQ for a consistent speaker, microphone and room, and bus EQ only for a broad change to a stem you already have under control.
A high-pass filter can take out rumble, or it can hollow out a low voice. The curve shows what you set; it can't tell you whether the dialogue still sounds believable.
Clip EQ vs Track EQ in DaVinci Resolve
| Placement | Use it for | Avoid it when |
|---|---|---|
| Clip EQ | One take, one microphone change, or one edit that differs from the track | Dozens of clips need exactly the same baseline curve |
| Track channel EQ | A consistent speaker, microphone, and room across many clips | The track mixes lav, boom, phone, ADR, and different locations |
| Bus EQ | A gentle broad adjustment to a controlled dialogue stem | Individual sources still have different defects |
| Automation or dynamic EQ plug-in | A frequency problem appears only on certain words or moments | A static, subtle cut already solves it |
For one clip, select it and go to Inspector, Audio, Equalizer. For a whole track, double-click the EQ area of its channel in the Fairlight Mixer. Blackmagic's current Fairlight product page describes six-band EQ on clips as well as on Mixer channels. A channel EQ affects every clip routed through that track, not just the one you selected.
The September 2026 Resolve 21.1 Reference Manual documents six-band clip EQ on pages 930–931 and six-band channel EQ on pages 3949–3952. Its Fairlight Inspector section, though, still says four-band on page 3912. That line contradicts the current product page and the six-band clip description in Blackmagic's training book, so don't assume the Inspector is limited to four bands.
The official Fairlight dialogue EQ lesson teaches clip, track and bus EQ on specific recordings. The book linked from Blackmagic's training page is still the Resolve 20 edition, while the rolling reference manual covers 21.1. Use its exercises to learn what to listen for, not as a curve to paste onto every voice.
In the channel EQ controls, Band 1 gives you a high-pass filter (low cut) and Band 6 a low-pass filter (high cut). High-pass reduces lows; low-pass reduces highs. Both outer bands also offer bell and shelving shapes. Bands 2–5 offer shelves, a peaking/bell curve and notch. A shelf moves a broad end of the spectrum, a bell boosts or cuts around a center, and a notch rejects a narrow region. The controls you see depend on the selected shape, so not every band shows Gain, Q and slope at once.
On a bell, lower Q widens the affected range and higher Q narrows it. Gain sets the boost or cut, while filter slope sets how quickly a high- or low-pass filter rolls off. The channel manual lists 6–24 dB/octave slopes, ±20 dB bell/shelf gain, and Q 0.3–10.3 where available. Those are control limits, not dialogue settings. It also documents Bands 1 and 6 as off by default and Bands 2–5 as on, so check a reused track for saved settings; don't assume it's flat.
If very different recordings share one track, split them out before you equalize anything. The organization steps in our Fairlight routing guide make later EQ, compression, automation and delivery much easier to verify.
Dialogue Frequency Ranges: Starting Clues, Not Presets
| Approximate area | What you might hear | Possible move |
|---|---|---|
| Below 60–100 Hz | Handling noise, traffic vibration, plosives, HVAC rumble (clean the dialogue first) | High-pass only until the problem improves |
| 200–500 Hz | Warmth or mud, depending on voice and room | Small broad cut if congested; do not remove body blindly |
| 250–750 Hz | Mud, enclosure, room coloration | Search for a resonance, then use a modest cut |
| 600 Hz–1.1 kHz | Nasal or honky tone | Narrow-to-medium cut only after confirming it in context |
| 2–5 kHz | Intelligibility, presence, or harshness | Broad small boost for clarity or cut for harshness |
| 5–8 kHz | Sibilance, hiss, consonant detail | De-esser or automation for changing S sounds |
| 9–15 kHz | Air, brightness, or broadband hiss | Very gentle shelf; avoid lifting noise |
Use these overlapping areas as listening clues, not borders or prescribed cuts. The training book’s dialogue frequency guide places warmth and boxiness around 200–500 Hz, mud around 250–750 Hz, and sibilance around 5–8 kHz. Voices, microphones, proximity, rooms and monitoring can all move the problem. A preset that specifies a cut before anyone has heard the recording can't choose it for you.
A Repeatable Fairlight Dialogue EQ Workflow
- Choose a representative loop. Include vowels, consonants, a breath and some room tone, and mark the worst phrase so you can come back to it later.
- Set clip gain first. Big level differences bias your tonal judgment and change how later compression behaves.
- Listen in the mix. Solo is good for identifying a defect; music, effects and the other speakers decide whether it matters.
- Enable a high-pass filter only for audible low-frequency contamination. Raise the cutoff slowly, then back off when the voice starts losing weight.
- Find one resonance. Use a modest, temporary narrow boost at a comfortable monitoring level and sweep carefully. When the offending ring or boxiness jumps forward, turn that boost into a small cut, then lower Q to widen the cut if that sounds more natural.
- Reduce masking before boosting clarity. A modest low-mid cut often uncovers speech without dragging up hiss or sibilance.
- Add broad tone only if needed. A small shelf or presence move has to survive matched-bypass listening, or it goes.
- Check other words and edits. A curve that fixes one vowel can wreck another.
- Render a difficult passage. Listen away from the on-screen curve and judge the rendered audio on ordinary playback.
Remove the temporary search boost before you judge the final cut. Keep the monitoring level comfortable and the boost modest; a louder sweep isn't better evidence. Add another corrective band only when you can name the problem it solves.
Set the High-Pass Filter by Ear
There's no correct universal dialogue cutoff, and a speaker's gender doesn't pick one for you. Low voice fundamentals and useful body can overlap the region you want to remove. Start below the useful voice, raise the cutoff until the low disturbance improves, then retreat if the voice loses weight. Compare the lowest notes, not just an average phrase. If you can't hear any rumble, leave the filter off.
Check plosives on their own. A high-pass filter may soften the low-frequency burst, but a cutoff high enough to erase it completely can thin out the entire sentence. Clip-specific EQ, a short gain envelope or dedicated repair is often cleaner than sacrificing every vowel.
Cut Mud and Resonance Without Hollowing the Voice
Mud is usually broad congestion; resonance is a narrower ring, and they need different treatment. For broad buildup, start with a small, low-Q cut and listen in context. For a ring, hunt it down with a narrower Q, then reduce the gain and widen the final cut until it stops calling attention to itself.
If every clip seems to need a deep 400 Hz notch, question the room, the microphone or the track assignment. EQ can improve coloration, but it can't remove the time smear of strong reflections.
If the defect is steady hiss, hum or broadband noise rather than tone, run the background-noise cleanup workflow before you start cutting wide frequency ranges. Voice-restoration tools have their own limits, and our Voice Isolation vs Dialogue Leveler guide explains when separation or leveling is the actual task.
For electrical hum, identify the fundamental and its harmonics: a 50 Hz hum can include 100, 150 and 200 Hz, and a 60 Hz hum can include 120, 180 and 240 Hz. Those are multiples, not all octaves, which is why a single notch can leave the higher buzz behind. The De-Hummer controls target that structure and include Listen to Hum only so you can check what's being removed. Switch back to normal monitoring before you judge the result.
EQ changes frequency balance wherever you apply it; it doesn't separate speech from noise sharing those frequencies. Noise Reduction targets noise, Voice Isolation separates a voice from competing sound, and a gate lowers material below its threshold, so a gate can't remove noise under speech while it's open. Compression controls changing level, and a de-esser controls excessive sibilance. Choose the defect first, and don't stack every one of these tools on the track.
Presence, Harshness, and Sibilance
The 2–5 kHz area can bring speech forward or make a recording sound harsh. If the dialogue is buried under music, lower or automate the music first. A big presence boost eats headroom and can exaggerate consonants without solving the competing sound.
Sibilance changes word by word, so a static high-frequency cut is usually too blunt. Undo the clarity shelf, identify the S range, then reach for a de-esser or clip automation. Check T, F and breath detail afterward; strip out every bright consonant and speech turns dull and less intelligible.
EQ Before or After Voice Isolation and Compression?
This is the logic I start from:
- Use clip gain before any processors so they receive a workable level.
- Correct audible low rumble on the clip when it's needed, then reassess any later restoration or compression.
- Run Voice Isolation before final tonal shaping when it substantially changes the spectrum.
- Put corrective EQ before compression if an ugly resonance would otherwise trigger the compressor.
- Use gentle tonal EQ after restoration when you need to bring back body or clarity.
- Compress after the corrective work, then recheck tone, because compression changes spectral balance perceptually.
The documented channel signal path places Track FX ahead of ordinary inserts, Dynamics and EQ. The Effects Order menu only changes the order of those last three groups, and its default is FX, Dynamics, EQ. So moving channel EQ before Dynamics still doesn't move it ahead of Track FX such as Voice Isolation.
That list describes listening decisions; it doesn't claim the channel Order menu can arrange every processor. Clip EQ can correct an individual source before it ever reaches channel processing. Change one stage at a time, recheck Voice Isolation for artifacts, and use the dialogue compression guide when uneven levels, not tone, are the problem.
Match Two Microphones Without Chasing Identical Curves
- Align clip gain and compare the same speaker at the same performance energy.
- Use the cleaner, more natural microphone as the reference.
- Remove each microphone's obvious defect on its own.
- Make one or two broad tonal moves on the mismatched source.
- Cut between the mics in context and listen to both the voice and the room tone.
- Add consistent ambience and crossfades wherever the acoustic background changes.
EQ can pull timbre closer, but it can't make a distant boom share the perspective, noise and reflections of a close lav. A slightly imperfect but stable transition beats two aggressively processed voices that change rooms at every cut.
Fairlight EQ in Free vs Studio
Built-in Fairlight EQ is available in DaVinci Resolve Free and Studio. You don't need Studio or a third-party plug-in for ordinary high-pass, low-pass, bell, notch, shelf, frequency, gain and Q work. Voice Isolation is explicitly Studio-only; built-in EQ isn't. Studio changes access to specific AI and advanced tools, but it won't turn a copied curve into a correct listening decision.
Troubleshooting Dialogue EQ
The voice becomes thin
Lower the high-pass cutoff, ease off the low-mid cut and check at matched loudness. Restore the last move; don't add a fresh bass boost to compensate.
EQ sounds good solo but disappears in the mix
Make the final decision with music and effects playing. The voice may be masked rather than badly equalized, so lower or automate whatever is competing with it. If the track is silent or its meters stop moving, EQ isn't the fix; work through the no-audio routing and monitoring checklist instead.
Sibilance gets worse
Remove the high shelf or presence boost and use a de-esser or phrase automation. Don't darken every word just to repair a few S sounds.
Room tone changes at each cut
Use clip EQ or separate tracks for the mismatched microphone, then add consistent room tone and crossfades. Track-wide EQ can't reconcile two unrelated acoustic spaces.
The mix clips after an EQ boost
Reduce band gain or channel output, then inspect the next bus. Every boost consumes headroom. Verify the final program with our LUFS and true-peak checklist.
Loop the same phrase, keep your monitor volume fixed, and compensate with the channel's post-EQ Gain control so processed and bypassed speech land at similar perceived loudness. Then toggle the EQ off and on. A/B compares two stored settings where it's available; it isn't automatically an untreated-versus-treated test, and it won't match loudness for you. Use the EQ's own enable/bypass control, because Effects In does not bypass built-in channel EQ. Watch the meters for headroom as well as listening.
Check on headphones and speakers at a consistent, comfortable level, then listen in mono. If a voice turns hollow only when two microphones play together, check timing, polarity and channel mapping before you add EQ. Fairlight’s Phase Meter can help diagnose cancellation on a bus, but it won't tell you which microphone sounds natural. Compare each source alone and together, and don't invert polarity blindly or boost bass to hide a cancellation.
Final EQ Check
- Every enabled band solves a named, audible problem.
- The high-pass filter removes rumble without thinning the lowest useful voice notes.
- No temporary search boost is left behind.
- Bypass has been compared at matched loudness.
- Dialogue still works with music and effects playing.
- Every speaker, microphone change and edit boundary has been checked.
- The mix translates to speakers, headphones and a small mono device.
- A rendered test matches the timeline; the audio export workflow covers that verification.
Keep only the changes that improve the words in context. If the untreated voice sounds more natural at the same loudness, undo the EQ move instead of adding another band to compensate.
For the next stage, the Fairlight audio hub links out to routing, cleanup, compression, loudness, recording and export guides.
Fairlight Dialogue EQ FAQ
What frequencies make dialogue muddy?
Mud tends to build up in the low mids; Blackmagic's training guide lists 250–750 Hz as a possible muddy area. The exact buildup depends on the voice, microphone, proximity and room. Confirm what you hear, then try a small broad cut without removing useful body.
Where should I set a high-pass filter for speech?
There's no universal cutoff. Raise it only until audible rumble improves, then back down before the voice loses body. Check every speaker and microphone.
Should I boost 3 kHz for dialogue clarity?
Not automatically. The 2–5 kHz area can improve intelligibility or create harshness. Reduce masking first, then test a small broad move in the full mix.
Is EQ better on clips or tracks?
Use clip EQ for one take, microphone change or edit that differs. Use track EQ when the speaker, microphone and room stay consistent. Save bus EQ for a gentle broad change after the individual sources are under control.
Can EQ remove room echo?
No. It may tame a prominent ring, but reflections overlap the voice across time and frequency. A better microphone choice, editing, ambience, restoration or a rerecord may be what it takes.
Should EQ come before compression?
Corrective EQ often comes before compression so rumble or resonance doesn't trigger it, but Fairlight lets you reorder insert effects, Dynamics and EQ. The default track order is FX, Dynamics, EQ; compare on the actual source.
Does Fairlight EQ work in DaVinci Resolve Free?
Yes. Built-in Fairlight EQ works in Resolve Free and Studio. Voice Isolation is explicitly Studio-only, but ordinary high-pass, low-pass, bell, notch, shelf, frequency, gain and Q work doesn't require Studio.
