DaVinci Resolve has its own noise tools, and they are not one button. There is a repair plugin you tune yourself, an AI model that does the tuning for you, and a third plugin for the hum that comes off bad cabling. All three live in the same place, and each one is the right answer for a different recording.
This page is about the three of them: where they are, which controls actually change the result, and the point where no setting will save the take.
What you need
- Resolve 21, the version this reference manual covers. The controls below were read from that manual, not from memory.
- The clip on a timeline, in the Fairlight page. The Edit page has the same effects with the same controls.
- Headphones. Two of the controls below exist only to let you hear what you are about to throw away, and a speaker in a room will hide the difference.
Pick the tool before you touch a control
The three plugins answer three different problems. Choosing the wrong one is why noise removal gets blamed for wrecking a voice.
| What you hear | The plugin |
|---|---|
| A steady hiss, air conditioning, room tone, tape noise | Noise Reduction |
| A buzz at 50 or 60 Hz, plus its harmonics | De-Hummer |
| A fan, traffic, a cafe, music under the voice | Voice Isolation (Studio version only) |
Noise Reduction is the one to learn, because it is the plugin that hands you the controls and lets you decide how much to take, instead of one amount dial.
1. Apply Noise Reduction to the clip
Open the Fairlight page, select the clip, and open the Audio FX panel in the Effects Library. Drag Noise Reduction onto the clip in the Timeline.
A badge appears to the left of that clip’s name bar, so you can see at a glance which clips carry an effect. If you select several clips first, dragging the plugin onto any one of them applies it to all of them.
Two other ways to place it:
- On a track. Drag Noise Reduction onto the track header. It then processes everything on that track, which is what you want for a long interview recorded on one mic.
- In the Mixer. Click the plus button in that channel strip and pick the plugin from the menu, or drag it onto the channel.
Clip or track is a real choice, not a shortcut. On a clip, the plugin sits on that clip alone. On a track, it sits on the channel and treats every clip that reaches it.
2. Listen to the noise on its own
Before you change a value, turn on Listen to Noise Only at the top right of the plugin window. You now hear exactly what the plugin is removing, and nothing else.
This is the most useful control in the whole plugin. If what you hear is a smooth hiss, the detection is right. If you hear words, breaths or the tail of consonants in there, the plugin is eating the voice and the settings are too hard.
The window also shows a spectral view of the audio with a purple overlay marking the noise being targeted, and two meters, input on the left and output on the right, so you can see how much level the reduction is costing you.
3. Set the threshold and the smoothness
Resolve’s presets are the fastest way in. There are three: De-Hiss, De-Rumble, and De-Rumble and Hiss. Pick the closest one, then adjust from there. Turn Listen to Noise Only back off to hear the result.
The controls, in the order they matter:
- Threshold, in dB. This relates to the signal-to-noise ratio of the recording. A recording with a poor signal-to-noise ratio needs a higher threshold, and a higher threshold means more reduction. Raise it until the noise is gone, not until the voice changes.
- Sensitivity. Higher values exaggerate the detected noise profile. More noise comes out, and more of the dialogue you wanted to keep goes with it. This is the control that makes a voice sound thin.
- Ratio. Sets the attack time of the signal profile against the attack time of the noise profile. A faster ratio keeps the transients in speech more easily, at the cost of a less accurate speech profile.
- Attack, in ms. The manual describes this as mainly useful in Auto Speech mode, where it sets how quickly the noise profile updates. A low value follows noise that changes quickly. A high value suits noise that stays the same.
- Frequency Smoothing and Time Smoothing. These clean up harmonic ringing left behind after the noise is pulled out. If the voice sounds metallic, add a little of each.
- Dry/Wet. A mix control. Pulling it back from 100 percent is a simpler fix than over-tuning the detection, and it keeps some of the original texture.
- Level. A pre-gain from -6 to +18 dB, applied just before the dry/wet mix, so you can put back the level the reduction took away.
Work in that order and stop as soon as the noise is gone. Every extra step of reduction costs something in the voice.
4. Use the Edit page if you are already there
You do not have to switch pages. On the Edit page, drag Noise Reduction from the Audio FX panel of the Effects Library onto the clip. The clip turns a darker shade, which is how the timeline shows that an effect is on it.
Select the clip and open the Inspector. Under the Effects tab, in the Audio section, the plugin’s parameters are right there. For the full floating window with the graph and the presets, click the custom control button, the one to the right of the trash can.
5. Take the hum out with De-Hummer
Mains hum is a different animal from hiss. It is a tone at 50 or 60 cycles, plus harmonics at multiples of it, usually from a cable, a power supply or a grounding problem. Noise Reduction will smear it rather than remove it.
Add De-Hummer the same way. Turn on Listen to Hum only to confirm you have the right tone, then:
- Frequency. Pick the 50 Hz or 60 Hz radio button that matches your mains, or turn the knob to the exact pitch you hear.
- Amount. How much hum is pulled out.
- Slope. Sets the balance between the fundamental and its harmonics. A value of 0 leans on the fundamental, 0.5 extracts all harmonics up to the fourth equally, and 1.0 targets the higher partials.
6. Level it after the cleanup, not before
Noise reduction does not change your levels, and it usually takes a little level with the noise. Set the loudness once the noise is gone, so you are measuring the audio you will publish.
Right-click the cleaned clip and choose Normalize Audio Levels, then enter the target for where the file is going. The steps and the targets are in how to normalize audio in DaVinci Resolve, and the LUFS explainer lists what each platform expects.
If the words inside the take also move up and down, Dialogue Leveler is the next tool. It rides down the loud parts and lifts the soft ones, and it has a Background Reduction option that gently removes background noise on top of its own levelling.
When it won’t help
The noise sits in the same range as the voice
A steady hiss is easy because it does not look like speech to the detector. A room full of chatter is hard for the opposite reason: some of it is voice, and the plugin has no way to know which voice you want. Turning up Sensitivity just removes the wrong one faster. This is why heavy settings give you a voice that sounds like it came from a phone in a lift, and why that happens is worth knowing before you push the slider.
The noise changes all the time
Wind gusts, passing traffic and a door that opens halfway through a take keep changing the noise profile. The detection is always one step behind. Cut those seconds out of the take instead, or re-record them.
Music is playing under the dialogue
Music is not noise, and no threshold will separate it. Music Remixer, which is Studio version only, splits a music track into stems so you can lower the vocal that clashes. Dialogue Separator, also Studio only, gives you separate levels for dialogue, background and room sound, on mono sources.
The voice is quieter than the noise
If the noise floor is close to the speech, there is nothing left to detect. Removing it takes the voice with it. Record again with the mic closer and the gain lower. A take that is too quiet can be raised later, a take buried in noise cannot be recovered.
The waveform is already clipped
A flattened peak is damage, and no plugin rebuilds the top of that wave. Noise reduction makes the distortion cleaner sounding, not smaller.
If the recording needs more than the timeline can give it, take the sound out of the video and treat it as a file. Pulling the audio out with ffmpeg is one command, and it keeps the picture untouched.
Cleaning the audio outside Resolve
Resolve’s plugins are tied to a timeline, a session and a project. When you just need a clean file, Gleam Audio does the same job on the recording itself, offline, on Mac, Windows and Linux.
- Remove noise is tuned for speech, which the app’s own known limitations say plainly. Interviews, voiceover and podcast takes are what it is built for.
- Every tool takes video files too, mp4, mov, mkv, webm and avi, and returns audio. You do not have to strip the sound out first.
- Enhance levels a recording in one step and shows the before and after numbers, and the editor has Cut, Split, Join and Convert for the rest.
- Free for personal use up to 20 files a week. A licence for paid work is $10 for Basic, $15 for Plus or $20 for Max, which is $25 at full price. Version 2.2.0, November 2026.
The full export, clean and put-back workflow is in how to reduce background noise in a video.
Takeaways
- Noise Reduction is the Fairlight FX plugin for hiss, room tone and steady noise. De-Hummer handles mains hum. Voice Isolation, Studio only, handles fans, traffic and music.
- Apply the plugin to a clip or to a whole track header, then listen to the noise on its own before you touch a value.
- Threshold follows the signal-to-noise ratio, Sensitivity is the control that eats the voice, and Level is there to put back what the reduction took.
- Level the audio after the cleanup, not before it.
- When the noise moves, sits under the voice, or is louder than the voice, cut the take or record it again. No setting fixes those.