Quick answer: fix large clip-gain differences before compression, then test the loud, normal, and quiet phrases. Blackmagic lays out 2.5:1–3:1 as commonly used dialogue ratios, but threshold must follow the actual input. Tune attack for intact consonants, release for smooth recovery, and use the post-Dynamics Makeup Gain only to match bypass. If room noise breathes or the processed version merely sounds louder, back off.
What a Dialogue Compressor Does
A compressor reduces gain when the signal crosses a threshold. Ratio controls how strongly it reduces the portion above that threshold; Attack and Release control when reduction begins and recovers. The resulting smaller dynamic range can keep loud words from jumping out and make softer words easier to place in the mix.
It does not repair clipped audio, remove background noise, or know which whisper should remain quiet for the story. It also does not set LUFS by itself. Compression is one stage between editing and final measurement.
The final July 2026 DaVinci Resolve 21 Reference Manual notes that every Mixer channel has built-in Dynamics with separate Expander/Gate, Compressor, and Limiter modules. Double-click the channel-strip Dynamics indicator to open the window; the graph maps input on the X axis, output on the Y axis, threshold as a blue line, and the active response as the green curve.
Blackmagic’s Fairlight Audio Guide provides project media for learning the wider dialogue workflow. For version-sensitive controls and ranges, this article follows the final Resolve 21 manual above rather than a copied preset or an older tutorial screenshot.
Clip Gain, Track Compression, or Bus Compression?
| Stage | Use it for | Do not ask it to |
|---|---|---|
| Clip gain / automation | Large edit-to-edit jumps, one bad phrase, microphone changes | Automatically follow an entire performance |
| Track compressor | Remaining dynamics from one consistent speaker and microphone | Reconcile unrelated rooms and recording chains |
| Dialogue bus compressor | Very gentle cohesion after tracks are balanced | Repair one bad track without moving the others |
| Main limiter | Catch final peaks against a delivery ceiling | Create a good dialogue mix from an unbalanced session |
Move boom, lav, phone, ADR, and narrator recordings to sensible tracks before processing. If the signal path isn't obvious, follow our Fairlight bus-routing guide before touching Main.
Fairlight Compressor Controls
- Threshold: the level where compression begins. Its useful value depends completely on the signal arriving at the track.
- Ratio: how much the level above threshold is reduced. At 2:1, 2 dB above threshold becomes roughly 1 dB at the output before other controls.
- Attack: how quickly gain reduction engages. Too fast can dull consonants; too slow can let a shout or plosive through.
- Hold: how long reduction is held before release behavior takes over. It can keep the compressor from fluttering between closely spaced syllables.
- Release: how quickly gain returns. Too short pumps; too long keeps the next phrase unnecessarily suppressed.
- Knee: how gradually the compressor enters its ratio around threshold. A softer knee often makes ordinary dialogue control less obvious.
- Mix: blends compressed signal at 0 with the original uncompressed input at 100. The default is 0, meaning all compressed signal. Use it deliberately; don't mistake the scale direction for a conventional wet control.
- Makeup Gain: a post-Dynamics level control shared by the module, not a compressor threshold target. It raises voice and noise together; use it to level-match bypass, not to win the comparison.
- Gain Reduction meter: the amount the compressor is turning down at that moment. It helps diagnose behavior but doesn't judge sound quality.
Set a Fairlight Compressor for Dialogue
- Level extreme clips manually. Bring obviously mismatched lines near the track’s normal working level. Keep intentional whispers and shouts unless the brief says otherwise.
- Choose three test sections. Loop normal speech, the loudest useful phrase, and a quiet phrase with room tone.
- Open Fairlight Dynamics. Enable Compressor and temporarily bypass Gate, Expander, and Limiter so you hear one process.
- Start with a gentle ratio. Blackmagic identifies 2.5:1–3:1 as commonly used for dialogue. Treat it as an audition range, not a preset; the recording and threshold still control how much compression occurs.
- Lower Threshold. Stop when louder words show a few decibels of reduction while ordinary speech moves less.
- Set Attack. Resolve 21 defaults to 1.4 ms and allows 0–100 ms. Start from the documented default, then lengthen only when consonants keep their edge and shorten when peaks still distract. The recording decides the value; this guide doesn't present an untested timing preset.
- Set Release. Resolve 21 defaults to 93 ms and allows 50–4000 ms. Start from the documented default and follow the phrase rhythm: the response should recover smoothly rather than chatter or hold down the next line. This guide does not claim one universal speech-release range.
- Adjust Knee and Hold if needed. Use a softer transition and enough stability to avoid syllable-by-syllable movement.
- Match Makeup. Toggle bypass and match perceived loudness before deciding which version is better.
- Review the complete scene. Check every microphone change, edit, breath, pause, and downstream bus.
Attack, Release, and Threshold are source-dependent. The official ranges are wide, and threshold is especially non-portable because clip gain changes what reaches the detector. A current editor discussion of long dialogue with sentence-by-sentence level changes points in the same practical direction: manual phrase gain before track compression is often cleaner than forcing very quiet lines through an extreme threshold. Community experience is directional; the manual controls the parameter facts.
How Much Gain Reduction?
For transparent ordinary dialogue, a few decibels on louder words is a useful place to begin. If gain reduction stays pinned through normal speech, ask whether the clip needs manual leveling or whether the threshold is too low. If only one peak triggers reduction, the compressor may be acting as a loose limiter rather than shaping the performance.
Don't chase a fixed meter number. Listen for a stable sentence, intact consonants, believable room tone, and natural differences between emotion and distance. Then compare in the full mix. A 6 dB reduction can be correct on one shout and obvious on another.
EQ, Isolation, Leveler, and Compression Order
Use clip gain first. Remove low rumble or a strong resonance before compression so the unwanted energy doesn't drive gain reduction; our dialogue EQ workflow explains that corrective stage. If Voice Isolation substantially changes the spectrum, use it conservatively before final EQ and dynamics.
Dialogue Leveler already rides speech. Stacking aggressive Dialogue Leveler and aggressive compression commonly raises breaths and produces pumping. The Voice Isolation vs Dialogue Leveler comparison helps decide whether you need separation, automatic leveling, conventional compression, or manual gain.
When Automation Beats More Compression
If a speaker leans away for one sentence, raise that clip or phrase. If one laugh is too loud, lower it. If music masks only three lines, automate the music. Lowering Threshold until the compressor solves every exception usually flattens the normal material and lifts the noise floor.
Automation also preserves story dynamics. A whisper can remain a whisper while still being intelligible; a shout can feel larger without clipping. The compressor then handles smaller performance variation instead of doing editorial triage.
Dialogue Bus Compression and Limiting
Once separate dialogue tracks already match, a dialogue-bus compressor can add gentle cohesion. Use a low ratio, little gain reduction, and slower behavior than an emergency clip repair. Bypass it on the whole scene, not one line, to confirm it improves continuity.
A limiter on Main is a peak guardrail after the mix. It isn't a replacement for clip gain, track compression, or correct loudness. Resolve 21 describes it as a very-high-ratio ceiling controlled by Threshold, Attack, Hold, and Release; too much limiting can remove internal dynamics. Measure after every bus processor with our LUFS and true-peak guide.
Free vs Studio
The built-in Fairlight channel compressor is available in both DaVinci Resolve Free and Studio. Voice Isolation and some other AI repair tools require Studio, but ordinary clip gain, EQ, dynamics, routing, automation, and loudness monitoring don't.
Troubleshooting Dialogue Compression
Dialogue pumps between words
Lengthen Release, add appropriate Hold, reduce Ratio or gain reduction, and check whether a Gate or downstream Limiter is creating the movement.
Consonants sound dull
Slow Attack or reduce compression. Listen at matched output level; the bypassed signal shouldn't lose simply because it's quieter.
Room noise becomes louder
Reduce Makeup and compressor workload. Compression narrows the difference between voice and noise; it can't separate them. Repair or automate the source first.
Quiet lines remain too quiet
Raise those clips or phrases before compression. Driving Threshold lower until every whisper triggers reduction will usually flatten normal lines.
The bus clips after Makeup
Lower Makeup, inspect the dialogue bus and Main, and check downstream EQ boosts. Makeup should match bypass, not exceed it.
The export sounds different
Confirm the Deliver page renders the intended bus and channel layout, then reimport and listen. Our audio export checklist covers the verification.
Final Compressor Check
- Large edit-to-edit jumps are fixed before the compressor.
- Normal, loud, and quiet passages have all been tested.
- Gain reduction moves smoothly rather than staying pinned.
- Consonants retain their leading edge.
- Breaths and room noise don't surge.
- Makeup is level-matched to bypass.
- The track, dialogue bus, and Main retain headroom.
- The rendered file has been reimported and checked.
If you stop hearing the compressor and simply understand every line more easily, the settings are doing their job.
To place compression inside the complete signal path, use the DaVinci Resolve Fairlight audio hub alongside the contextual routing, EQ, Voice Isolation versus Dialogue Leveler, loudness, and export guides.
Fairlight Dialogue Compressor FAQ
What ratio should I use for dialogue?
Blackmagic walks through 2.5:1 to 3:1 as commonly used for dialogue processing. Treat that as an audition range, not a preset: incoming level, threshold, performance, and intended dynamics still decide the result.
How much gain reduction is enough?
A few decibels on louder words is a transparent starting point. If normal speech remains heavily reduced, fix clip gain or use less compression.
What attack time works for speech?
Resolve 21 defaults to 1.4 ms and allows 0–100 ms. Start from the documented default, then lengthen only while consonants keep their edge; the recording decides the value.
Why does compression raise background noise?
Gain reduction narrows the difference between voice and room sound, then the post-Dynamics Makeup Gain raises the complete result. Reduce compressor workload, match bypass level, and repair or automate the source first.
Should I use Dialogue Leveler and a compressor together?
Only if each solves a remaining audible problem. Dialogue Leveler already rides level, so keep both stages gentle and watch for breaths, noise, and pumping.
Should dialogue compression go on clips or tracks?
Correct exceptional clips with gain or automation. Use track compression for a consistent speaker and microphone, and bus compression only for subtle shared control.
Is the Fairlight compressor available in Resolve Free?
Yes. The built-in channel Dynamics compressor is part of the standard Fairlight toolset in Free and Studio.
Related guides
- Fairlight Audio in DaVinci Resolve: the hub: the dialogue chain this step belongs to.
- How to Remove Background Noise in DaVinci Resolve: compression raises noise; clean first.
- DaVinci Resolve Audio Ducking: sidechain dynamics between music and dialogue.
