Audio in the edit
Almost half of a video is sound, and the edit is where it shows whether someone knows what they're doing. Balanced levels, a clear voice, music that supports without drowning things out and no distracting noise: that's what makes a video “sound professional”.
Organize your tracks
Before you touch a single level, get organized: one track for dialogue, another for voice-over, another for music and another for sound effects and ambience. That way you can adjust each group separately and the mix will go much faster.
Loudness: LUFS
The peak meter (dBFS) tells you whether the sound is clipping. But to know whether a video “sounds loud” or “sounds quiet” you use integrated loudness, measured in LUFS: an average over the whole video weighted to match how the ear perceives sound.
| Destination | Rough loudness | Maximum true peak |
|---|---|---|
| YouTube and most web platforms | −14 LUFS | −1 dBTP |
| Podcast | −16 LUFS (stereo) | −1 dBTP |
| Broadcast TV: UK and Europe (EBU R128) / US (ATSC A/85) | −23 LUFS / −24 LKFS | −1 dBTP (−2 dBTP in the US) |
| Social media on mobile | −14 to −12 LUFS | −1 dBTP |
Platforms normalize volume: if you deliver a video that's much louder, they turn it down; if it's very quiet, it may stay quiet. Aim for the reference and don't worry about it any further.
In DaVinci Resolve, on the Fairlight page, the meters panel (Loudness) shows integrated LUFS. Play the whole video, or at least a representative stretch, to read the value.
Voice, music and effects: the hierarchy
The voice rules. As a guide when mixing:
- Dialogue: peaks around −6 / −10 dBFS, loudness close to the delivery target.
- Music under the voice: 15 to 25 dB below the voice. If in doubt, lower.
- Music on its own (no voice): it can come up almost to the level of the voice.
- Ambience and effects: felt rather than noticed.
Ducking: lowering the music when someone speaks
Ducking automatically lowers the music when there's voice and brings it back up in the pauses. You can do it:
- By hand with volume keyframes: more control.
- With a compressor on a sidechain: the voice controls the music's volume.
- With the automatic features in many editors.
The music should start dipping a few frames before the voice comes in and rise gently when it ends. If it drops suddenly right as someone starts talking, you notice the “hole”.
Cleaning up the voice, in this order
- Noise reduction: in moderation. Overdo it and the voice sounds metallic, “underwater”. Some AI-based tools isolate the voice very effectively (see AI for video editing).
- EQ:
- High-pass filter at 80–100 Hz: removes rumble and thumps without touching the voice.
- If it sounds boxy, cut a little at 200–400 Hz.
- For more clarity, a gentle boost at 3–5 kHz.
- If the s sounds hiss, a de-esser at 5–8 kHz.
- Compression: evens out the loud and quiet parts. To start: 3:1 ratio, medium attack and 3 to 6 dB of gain reduction on the peaks.
- Limiter at the end of the master: ceiling at −1 dB so nothing clips.
If the music is covering the voice, don't push the voice up until it clips: turn the music down. And if the voice still isn't clear with the music lower, the problem may be EQ: the music is competing in the voice's frequencies.
Listen in several places
Mix on headphones or monitors, but check on a phone speaker and a laptop speaker: that's where most videos will be watched. If the voice isn't clear there, your mix doesn't work.
- One track per type of sound.
- −14 LUFS for the web; −23 LUFS for UK/European TV (−24 LKFS in the US).
- Music 15–25 dB below the voice, with gentle ducking.
- Noise reduction, EQ, compressor and limiter, in that order.
- Check on a phone speaker.
Your turn: a one-minute mix
Take a minute of an interview or of your own voice and add royalty-free music.
- Clean up the voice: high-pass, gentle compressor.
- Duck the music by hand with keyframes.
- Measure the LUFS and set the overall level to −14. Listen on a phone speaker.