How to add music to a video: timing, ducking, licensing and where to get the track

Adding music to a video takes four decisions and about ten minutes: which track, where it starts, how loud it sits under speech, and how it ends. The mechanical part is a drag onto an audio track in any timeline editor. The part that separates a professional edit from a slideshow is timing the cuts to the beat, ducking the music under dialogue, and choosing a track you are actually allowed to publish. This guide walks through all four in a browser-based timeline, with the music generated to fit rather than trimmed to fit.

What do you need before adding music to a video?

You need the finished picture edit, a music file you have the right to use (your own generation, a licensed track or a public-domain recording), and a timeline editor with at least one separate audio track and volume keyframes.

Add music last. Every cut you make after the music is placed breaks the sync you just built, so lock the picture first. The track itself can come from three places: a library license, a recording that is genuinely public domain, or a song generated for the video. Generated music has one practical advantage over the other two: you can ask for the exact length, tempo and mood the cut needs instead of forcing the cut to fit a three-minute song. PrismPoster's AI music generator makes a custom track with vocals in 35 languages for 25 credits, or an instrumental bed, and delivers stems so the vocal and the instrumental can be placed separately.

Step 1: place the track on its own audio track

In the timeline editor, drop the music onto an audio track below the video, not onto the video clip itself. A separate track is what lets you trim, fade and duck the music without touching the picture. Line the start of the track with the first frame that should have sound, which is usually not frame one: an opening title with a second of silence before the music lands reads as intentional, while music that starts on frame zero reads as default.

Step 2: cut to the beat, or move the beat to the cut

Music sounds "right" under a video when the picture cuts land on downbeats; you achieve that either by nudging cuts to the beat grid or by starting the track a fraction earlier so a downbeat falls on an existing cut.

Find the tempo first. Tap along with the BPM tapper for ten beats and it returns the BPM plus the length of one bar; at 120 BPM a beat is 0.5 seconds and a 4/4 bar is 2 seconds. Then either snap your cuts to those intervals or slide the track so that a strong downbeat hits your most important cut. You do not need every cut on a beat; the reveal, the title and the last shot are enough for the whole edit to feel locked.

Step 3: duck the music under speech

When someone talks, the music drops; when they stop, it comes back. That is ducking, and it is what keeps a voice-over intelligible without turning the music off. The numbers that work for most speech: music at full level between lines, down by 12 to 18 dB under dialogue, with a fade of about a quarter of a second on each side so the change is not audible as a jump. Use volume keyframes on the music track, placed a beat before the line starts and a beat after it ends. If the music has its own vocals, duck harder or switch to the instrumental stem under any spoken line; two voices at once is the one mix nobody can follow.

Step 4: end the music on purpose

Three endings work. A fade-out over two to four seconds is the safe default and hides any track that runs long. A hard out on the last cut lands when the song's phrase ends there, which you can arrange by trimming the track from the front so its final downbeat meets your final frame. A tail, where the music runs a second past the last shot over black, is the most cinematic and needs a track whose real ending you can use. Generated music helps here as well: ask for a track of the runtime you need and the ending is composed rather than cut.

Which music are you allowed to use?

A track is safe to publish only if you hold a licence for it, it is in the public domain, or you generated it under terms that grant you the rights; "royalty-free" is a licence type, not a synonym for free, and a copyright claim on a platform is enforced against the video, not the person who made the mistake.

Library music comes with a licence document; read the part about where the video may be published and whether monetisation is allowed. Public-domain recordings are rarer than they look, because a public-domain composition can still have a copyrighted recording. Generated music follows the generator's terms: what PrismPoster grants, and where the limits are, is set out plainly in AI music licensing explained, and the platform-specific question is answered in can you use AI music on YouTube. Keep the stems and the generation record with the project; if a platform ever asks where the music came from, that is the answer.

How do you add music without covering the original sound?

Keep the video's own audio on its track and mix rather than replace. Location sound (footsteps, room tone, a door) is what makes a scene feel real; music alone makes it feel like a montage. A typical mix has the location sound at full level, music 10 to 15 dB below it, and both ducked under speech. If the original audio has a song in it that you cannot license, remove it: the free vocal remover separates a song into vocal and instrumental stems in your browser for your own recordings, but a copyrighted song stays copyrighted after separation, so the only clean fix for someone else's music in your footage is to replace it.

How long should a music bed be?

As long as the video, plus the tail. A social clip of 4, 6, 8, 10 seconds rendered in the Video Studio can take a bed generated to that exact length; a 60-second explainer usually wants a track that is 62 to 64 seconds so the fade lands over the end card. Generating to length is cheaper than editing a long song down, because a song has a structure (intro, verse, chorus) that a random 60-second slice cuts through the middle of. If you are building the whole piece, how to make a music video with AI covers the reverse problem, cutting picture to a finished song, and the browser video editor overview covers the timeline itself.

If music is the part you are working on, these go further: Free vocal removers compared, How to choose an AI music generator, writing song lyrics with AI, How to make AI music for free, and How to make a song with AI.

Frequently Asked Questions

How do I add music to a video for free?

Any browser timeline with an audio track can do it, including PrismPoster's editor on a free account, which starts with 150 credits in most regions. The music itself is the cost: a licensed library track, a public-domain recording, or a generated track at 25 credits for a custom song.

How loud should background music be under a voice?

About 12 to 18 dB below the voice while someone is speaking, returning to full level between lines, with quarter-second fades so the change is not audible as a step. Instrumental beds can sit slightly louder than tracks with vocals.

Can I use any song on YouTube if I credit the artist?

No. Credit is not a licence. A song needs a licence from its rights holder, or it needs to be public domain or generated under terms that grant you the rights; otherwise the video can be claimed, muted or blocked. See AI music licensing explained.

How do I make the music match the cuts?

Find the tempo with a BPM tapper, then slide the track so a downbeat lands on your most important cut, or nudge the cuts onto the beat grid. Locking three or four key cuts is enough for the whole edit to feel in time.