To add background music to a video with AI, you don't dig through a stock library hoping to license a clip that fits, you generate an original score that matches your video's length, mood, and pacing, then balance it under your voiceover on a timeline. CoreReflex does this with Lyria for generation and a real audio mix on the editing timeline, this walkthrough covers generating the score, ducking it under narration, matching length, and balancing the final mix.
What you need before you start
Background music is the last layer, not the first. Have your visuals cut and your voiceover (if any) placed on the timeline before you score, because the music should follow the edit's rhythm, not the other way around. You'll want a rough sense of the mood you're after (upbeat, tense, calm, cinematic) and the video's total runtime, since both feed the generation.
If your video doesn't have narration yet, that's fine, instrumental background music works on its own. But if you plan to add a voice, lay it down first so you can duck the music against it. CoreReflex's voiceover seam can generate that narration with a named HD voice, so the whole soundtrack, score and voice, comes from one pipeline.
How to add background music to a video with AI
Here's the end-to-end path inside CoreReflex:
- Open your project on the timeline. Place your video clips and any voiceover track first so you can see the structure you're scoring against.
- Generate a score with Lyria. Describe the mood, genre, and energy you want, for example, "warm, optimistic acoustic with a gentle build." Lyria generates an original track rather than pulling from a stock catalog, so the music is made for your project.
- Drop the music onto an audio track. Add it as its own layer beneath your video and voice, so each element stays independently adjustable.
- Set the music level low to start. Background music belongs in the background. Begin well under your voiceover so dialogue stays the focus, then fine-tune.
- Duck the music under narration. Lower the music wherever the voice speaks so the words stay clear (covered in detail below).
- Match the music length to the cut. Trim, loop, or regenerate so the score resolves cleanly at the end instead of cutting off mid-phrase.
- Balance the full mix and render. Adjust each track's level, then render on the deterministic path so the same manifest always produces the same mixed output.
Because every generation in CoreReflex carries a portable provenance trace, model, prompt, and parameters, you can reproduce or tweak the exact score later instead of trying to remember what you typed. If you want to go deeper on the generation side, our guide to generating original music for a video covers prompting Lyria for specific moods and structures.
Duck the music under your voiceover
Ducking is the single most important move for any video with narration. It means automatically lowering the music whenever the voice is speaking, then bringing it back up in the gaps. Done right, it's invisible, viewers just notice that they can hear every word and that the music "breathes" around the narration.
On the CoreReflex timeline, you keep music and voice on separate audio tracks so they can move independently. Set the music a comfortable distance below the voice during speech and let it rise during pauses, the intro, and the outro. The goal is a mix where the voice always wins a head-to-head but the music never disappears entirely.
How much should you duck?
A practical target is to drop the music noticeably during speech and restore it in the spaces between lines. Trust your ears over any fixed number: if you find yourself straining to catch a word, the music is too loud; if the track vanishes the moment anyone speaks, you've gone too far. The voiceover seam and the music generator both run through the same project, so once your narration is placed, balancing the two is a timeline task, not a re-export.
Match music length to your video
Nothing gives away a rushed edit like music that stops abruptly two seconds before the video ends, or loops awkwardly back to the top mid-scene. You have three ways to fit the score to your runtime:
- Trim to a natural resolution. Cut the track at a musical phrase boundary so it feels intentional, then fade the last beat into your outro.
- Loop seamlessly. For longer videos, loop a section that's built to repeat, hiding the seam under a cut or a louder visual moment.
- Regenerate to length. Because the score is generated, the cleanest option is often to generate again with the target duration in mind so the music builds and resolves to fit the cut.
Matching music to picture is closely related to pacing the whole edit. The same instinct that keeps a score from running long keeps your visuals from dragging, a discipline that pays off across formats, from 30-second commercial spots to longer explainers.
Balance the final mix
With music ducked and length matched, the last step is the overall balance. Listen to the whole video front to back, ideally on the kind of speakers your audience will use, most social video is watched on phone speakers, so check there. Confirm three things: the voice is always intelligible, the music supports the mood without competing, and there are no sudden jumps in loudness between sections.
Then render. CoreReflex's render-worker produces a deterministic, faststart-encoded file, so the mix you approved is exactly the mix that ships, and the same project re-renders identically every time. Once the audio is right, you can layer on the finishing touches that lift retention, like syncing captions to your video so the message lands even when the sound is off, or adding animated captions for extra punch on social. For more timeline and editing walkthroughs, browse the full AI video hub.
Frequently asked questions
Is the background music royalty-free?
The score is generated by Lyria specifically for your project rather than pulled from a stock library, so you're not matching your clip against a third-party music license catalog the way you would with a downloaded track. Every generation also carries a provenance trace recording the model and parameters used. As with any AI-generated asset, review the applicable platform terms for your intended use rather than assuming a blanket license.
Can the music duck under my voiceover automatically?
Yes, keep music and narration on separate audio tracks and lower the music wherever the voice speaks, raising it in the gaps. Because the voiceover seam and Lyria run through the same project, you set this balance directly on the timeline without re-exporting. Aim for a mix where the voice always reads clearly but the music never fully disappears.
How do I match music length to my video?
You have three options: trim the track to resolve at a musical phrase boundary, loop a repeatable section for longer videos, or regenerate the score with your target duration in mind. Because the music is generated rather than licensed, regenerating to length is often the cleanest way to get a score that builds and resolves to fit the cut exactly.
Should I add music before or after my voiceover?
Add the voiceover first. The music is the supporting layer, so you want to score against the finished narration and visuals, then duck and balance the music underneath. Building the score first forces you to fight the edit's rhythm instead of following it.
Score your video and ship the mix
Generate an original Lyria track, duck it under your narration, match it to your runtime, and balance the whole thing on the timeline, that's the complete path to background music that sounds produced rather than pasted on. Because the render path is deterministic, the mix you approve is the mix that ships. Start free with no credit card and score your next video in a single session.