To add music to a video on iPhone, open your editor's audio tools, add a music bed, then set it to duck under your voice automatically. Ducking is the step most people skip, and a bed at a fixed level fights your narration for the whole video. Aim for music at roughly 15-25% of your voice level, which is about 12 to 16 dB down, and fade it out at the end rather than cutting it dead.
To add music to a video on iPhone, open your editor's audio tools, drop in a music bed, and then - the step that actually matters - set it to duck under your voice. A track at a fixed level competes with your narration for the entire video; a ducked track drops while you speak and swells back in the gaps.
What the built-in apps can and cannot do
The Photos app cannot add music. Its video editor trims, crops, rotates and adjusts a clip you already have. There is no audio track to add to, and the only audio control is the speaker icon in the top-left of the edit view, which mutes the whole clip.
iMovie can, with one significant wall. Tap the + button on the timeline and choose Audio, and you get three sources: Soundtracks (iMovie's own themed loops, which automatically fit themselves to your project's length), My Music, and Sound Effects. The wall is My Music: anything downloaded through an Apple Music subscription is DRM-protected, so it appears greyed out and refuses to import. Only tracks you actually own as files will go in. This surprises almost everyone, and it is not a bug you can work around.
Adding music inside TikTok or Instagram is a different thing entirely. Their catalogues are licensed for use on that platform. A sound you add inside the TikTok editor covers the TikTok post; take the same video to YouTube and that licence does not travel with it. Baking music into the file yourself is the portable option, and it is on you to have the rights.
For the underlying concepts, Apple's audio fundamentals is a clear primer on levels, headroom, and why a mix that peaks correctly can still sound wrong.
On iPhone and iPad, screen capture goes through ReplayKit, Apple's system framework for recording and broadcasting the screen.
The two-minute version
- Add the bed. Use a built-in loop or import a track you own. It should start at the top of your video, not wherever the playhead happens to sit.
- Turn on ducking. The bed dips automatically while you speak. Without this you will spend the rest of the edit keyframing volume by hand.
- Set the base level. Music sits somewhere around 15-25% of your voice. If you notice the music while someone is talking, it's too loud.
- Check the ending. A track that stops dead at the last frame sounds like a mistake. Fade the last second or two, or let it resolve after your last word.
- Listen on the target device. Phone speakers flatter music and bury speech. Check on headphones, then on a laptop speaker.
Levels, in numbers you can actually aim at
"15-25% of your voice" is a slider position, and it maps to something concrete: 20% amplitude is about 14 dB below the voice, 25% is about 12 dB down, and 15% is about 16 dB down. So the working range is roughly 12 to 16 dB under dialogue, which is where broadcast mixes tend to sit too.
Two things make that range work rather than a specific number:
- Normalise the voice first. If your narration is normalised to a known loudness (around -14 LUFS integrated is a sensible target for social video), then a percentage of it means something. If the voice level wanders, the bed will feel right in one section and wrong in the next.
- Judge it by the sentence, not the meter. Play a full sentence and ask whether you noticed the music. If you did, drop it 5%. That test is more reliable than any number, including these.
Why ducking matters more than the track you pick
Viewers forgive an ordinary music choice. They don't forgive not being able to hear you. Ducking (also called sidechaining) is what makes a bed feel intentional: the music has room to breathe in your pauses and gets out of the way the moment you speak.
Doing it manually means keyframing the volume down at the start of every sentence and up at every gap. For a 60-second video that is dozens of keyframes, and every re-cut invalidates them. Automatic ducking derives the same envelope from the speech itself, so it survives editing.
Two things to know about how ducking behaves:
- It triggers on sound, not on meaning. A cough, a laugh or a loud door will duck the bed too. That is usually fine and occasionally distracting.
- The recovery matters more than the dip. A bed that snaps back to full volume the instant you stop talking sounds like it is breathing at you. A recovery spread over a fraction of a second reads as a deliberate mix.
Where the music starts and stops
- Start it on the take's first beat, not at frame zero of a silent title card, unless you want the card to feel scored.
- Do not let it die at a hard cut. If your video ends with an outro card, the bed should carry across the cut and fade out on the card, or the card lands in silence and sounds broken.
- Trim the bed to the video, then fade. A one to two second fade at the end is the difference between "the video ended" and "the file stopped".
Licensing, briefly
If you plan to post publicly, use music you're allowed to use: built-in loops that ship with your editor, a licensed library, or a platform's own audio catalogue with the platform's own limits in mind. "It was in another video" is not a licence, and audio matching on TikTok, Reels and YouTube is very good. The realistic downside is not a lawsuit; it is your video being muted or demonetised after it starts working.
How Zella does it
In Zella for iPhone & iPad, tap Audio in the timeline toolbar. That one sheet holds:
- A music bed with Auto-Duck - it dips under your voice and swells in the gaps, no keyframing.
- Clip volume (0-200%) and mute, plus a level for every picture-in-picture overlay.
- 30 sound effects, 16 voice changers and five voice tones.
- Polish Voice - on-device noise removal, de-ess, compression and -14 LUFS loudness, which is what actually makes narration sit "on top" of a bed.
Auto-Polish also lays a mood-matched bed automatically: it picks a loop to suit your pace, starts it on the take's first beat, ducks it under speech and fades it out at the end. Beat Sync is there if you want cuts and punches landing on the kick or the snare, and it is never applied for you.
Everything runs on the device - no upload, no account, and the whole audio toolset is free. The Mac app is on the Mac App Store today; Zella for iPhone and iPad is coming to the App Store.
FAQ
Can I use a song from Apple Music? No. Apple Music downloads are DRM-protected and cannot be imported into iMovie or any other editor. You need a file you own, a licensed library track, or a built-in loop.
Can I add music to only part of the video? Yes - trim the bed on the timeline so it covers just the section you want, or add it where you need it rather than at the start.
Why does my music get quiet in random places? That's ducking doing its job: it responds to speech, and to coughs, laughs and background noise. If it feels twitchy, lower the base level so the dips are less dramatic.
Should I add music in the editor or in the TikTok app? Adding it in-app links your video to that sound's page, which can help discovery, but the licence covers that platform only. Baking it in travels with the file and works everywhere you post it.
How loud should music be under narration? Start around 20% of your voice level, roughly 14 dB down. The test is simple: play a sentence and ask whether you notice the music. If you do, it's too loud.
Make your next video with Zella.
Record, edit and ship on Mac, iPhone and iPad - local, private, free to start.
RELATED