Why video audio is so often noisy
A camera or a phone records sound from where it stands, not from where the person speaking is. The further the microphone is from the voice, the more of everything else it picks up: traffic, air conditioning, the room’s echo, wind across the microphone outdoors.
None of that can be filtered out of the picture, but most of it can be reduced in the sound, as long as you treat the right problem. Hiss, hum, wind and echo are different problems, and the rest of this guide takes them one at a time.
The method that works with any editor
Whatever you edit in, the dependable route is the same three steps:
- Export the audio on its own, as a WAV file, from the very start of the timeline to the very end, at the project’s sample rate. Most video projects run at 48 kHz.
- Clean that file with a tool built for the noise you have.
- Import the cleaned file, place it at the start of the timeline, and mute the original audio track.
It takes a few minutes longer than a button inside the editor, and it works the same way in Premiere Pro, DaVinci Resolve, Final Cut Pro, CapCut or anything else that can export audio and import it again.
Getting the audio out
Every editor can export audio only. Look for an export setting where you can switch video off, or choose an audio format, and pick WAV. Three settings matter more than the menu they live in:
- WAV, not MP3 or AAC. Lossy encoders add a short delay at the start of the file, so a round trip through them can shift the sound slightly against the picture. WAV keeps every sample where it was.
- The project’s sample rate. Exporting at a different rate makes the editor convert the file on the way out and again on the way back in.
- The whole timeline, from the first frame. If the exported file starts where the timeline starts, putting it back is a matter of dropping it at zero.
Cleaning it
Listen to a few seconds where nobody is speaking and work out what is actually there. A steady hiss or rush is broadband noise. A low tone with a buzzy edge is mains hum. Clicks are damage, and a voice that sounds distant or hollow is the room. How to remove noise from a recording explains each one and why one blanket filter makes voices sound underwater.
Euphona’s audio restoration treats them separately: noise reduction, hum removal, declicking, declipping and dereverb, each one you can switch on or off, with one strength control. It works on audio files, which is why the audio comes out of the video first. You hear the result free, matched in level against the original, and buy the full-quality file only if it is better.
Whatever you use, use less than you want to. Background noise reduced until it stops being distracting sounds natural. Background noise removed completely usually takes part of the voice with it.
Putting it back in sync
Import the cleaned file, place it at the very start of the timeline, and mute or delete the original audio. If the export started at the first frame, it lines up with nothing else to do.
That depends on the cleaning step leaving the file exactly as long as it was. Euphona’s restoration does: we measured it on files at 48 kHz and 44.1 kHz, with every repair switched on, and the output had the same number of samples, the same sample rate and the same channels as the input. A tool that trims silence or changes the sample rate would break the sync, so check that first if a file comes back shorter.
A quick way to confirm the alignment: find a hard consonant like a “p” or a “t”, zoom in on the waveform, and check the cleaned track and the original start it at the same moment.
What your editor already has
Some editors now include a one-button speech cleaner, and for a talking head with steady background noise it is worth trying first.
- Premiere Pro has Enhance Speech in the Essential Sound panel, with a Mix Amount slider that blends the enhanced and original sound. Adobe describes it as a tool for speech that will not work on music, sound effects or ambient sound. Adobe Premiere Pro help, “Enhance speech”, and Adobe’s announcement of the feature.
- DaVinci Resolve has Voice Isolation, which Blackmagic Design lists among the features of the paid DaVinci Resolve Studio rather than the free version. Blackmagic Design, DaVinci Resolve Studio product page.
These tools are built to keep a voice and remove everything else, so they work best when speech is all you want. When the background is part of the scene, like music in a vlog, the sound of an event, or an interview where the room matters, a tool that repairs specific problems keeps more of what you want: it removes the hum or the hiss and leaves the music and the room alone.
Wind noise
Wind is the hardest noise in outdoor video, because it is not steady. It arrives as low rumble and sudden bursts, and a strong gust can overload the microphone completely.
- Start with a high-pass filter. Most wind energy sits in the low frequencies, below the voice. Cutting the lowest range removes much of the rumble with no side effects.
- Then reduce what is left, gently. Noise reduction handles the steadier part of the wind. The bursts change too quickly for it to follow.
- Repair overloaded gusts last. Where a gust pushed the recording into clipping, declipping can rebuild part of the damaged waveform, but not all of it. Those moments may be better covered with music or cut.
Next time, a foam or furry windscreen on the microphone prevents more wind noise than any processing removes.
Echo and room sound
A voice that sounds hollow or far away was recorded in a reflective room. That is not noise in the usual sense: it is the voice itself, arriving again from the walls. Noise reduction will not touch it. Dereverb will, by reducing the reflections and keeping the direct sound.
Reduce the room until it stops being distracting, not until it is gone. A voice with no room at all sounds wrong to listeners, even when they cannot say why.
Next time, before you record
Move the microphone closer to the voice, or use a clip-on microphone. That raises the voice against everything else and is the single most effective thing you can do. Switch off fans and air conditioning for the take. And record ten seconds of the room with nobody talking: it gives any noise reduction tool a clean sample of exactly what to remove.
The questions people actually ask
Can I remove background noise from a video for free?
You can try it for free. With Euphona you export the audio, upload it and hear the cleaned result at no cost, matched in level against the original; the full-quality file is a paid download. The speech tools inside editors come with the editor, and in DaVinci Resolve’s case that means the paid Studio version.
Why does my voice sound robotic after noise removal?
That watery, warbling sound has a name, musical noise, and it means more reduction was applied than the recording supports. Reduce less, or give the tool a cleaner sample of the noise. A little audible background sounds natural; the warble does not.
Will the cleaned audio still match the lips?
Yes, if the export starts at the first frame, stays in WAV at the project’s sample rate, and the cleaning tool keeps the file’s length. Euphona’s restoration keeps it to the sample.
Can I upload the video file itself?
Not to Euphona: it takes audio files, in WAV, FLAC, MP3, M4A, OGG or AIFF. Export the audio from your editor first, as above.
Keep reading