Add Background Music to Video with AI
To add background music to a video, bring the track (upload it, paste a link or name the song) and describe the mood you want. Valmera’s AI agent fits the track to the video’s mood and pacing: it measures the tempo, beats and energy, lands the drop on the key moment, ducks the music under speech, fades it in and out, and can cut the edit to the beat.
Valmera doesn’t generate music or pick tracks from a library: you choose the song, and the agent does the sound editor’s part.
Fit Your Track to Your Video
Upload the video and the song, describe the feel, review the preview. Creating an account and uploading are free; editing requires a subscription.
Create an account →How to Add Background Music to a Video in Valmera
- 1Upload the videoMP4, MOV, M4V, MKV or WebM, up to 14 GB and 3 hours. Valmera indexes it once, so the agent knows where people speak and where the picture peaks.
- 2Bring the trackUpload an MP3, WAV, M4A, AAC or OGG file up to 50 MB, paste a YouTube, SoundCloud or Dropbox link, name a specific song, or point at a video clip whose audio you want.
- 3Say what the music is for“A quiet bed under my talking.” “Start at the chorus.” “Land the drop on the reveal at 0:42.” With no other instructions, the music runs under the whole video with context-aware levels and fades.
- 4Review and exportThe agent renders a preview and checks the mix before it replies. Ask for changes in plain English (“quieter under my voice,” “try the other song”), then click Export in Studio for a full-resolution MP4 rendered from your original file.
Add this track under the whole video as a quiet bed and fade it out over the last 3 seconds.Use the audio from the clip I uploaded and land the drop on the reveal at 0:42.Find “[song] by [artist]”, start it at the chorus and cut the montage to the beat.Remove the music that was in my original recording but keep my voice.Match the Music to the Video
Pick a track for what the video is, then for its mood. The tempo should roughly match the cutting pace, and under speech an instrumental works better than a song with vocals, which compete with the words.
| Video | Music that fits | Tempo | Level in Valmera |
|---|---|---|---|
| Talking head, podcast, tutorial | An unobtrusive instrumental | Anything that stays out of the way | −18 dB bed, ducked under speech (the default) |
| Tech or product demo | Clean, minimal electronic | Steady | Bed under the narration |
| Vlog | Lofi or indie warmth | Relaxed | Bed that swells back between lines |
| Emotional story | Sparse piano or ambient | Slow | Low bed with long fades |
| Fast montage, gameplay, sports | Driving and rhythmic | 120–160 BPM | Lead at −4 dB, cuts on the beat |
| Cinematic reveal or trailer | A track that builds to a drop | 60–90 BPM | Lead, with the drop on the key moment |
How Valmera Fits the Track
- It measures the track. Valmera reads the song’s tempo, beats, loudest and quietest sections and biggest energy rise, and the agent can listen to excerpts of the song and the rendered mix to judge the fit against the mood you described.
- It lands the drop on the moment. The agent finds the video’s peak (the reveal, the punchline, the hardest cut) and offsets the song so its build leads into it: if the drop is 48 seconds into the song and the reveal is at 0:20, the music starts 28 seconds in. Slow intros get skipped the same way.
- It sets the level by context. Where speech survives under the music, the track becomes a bed at −18 dB with smooth sidechain ducking: it dips with the voice and swells back in the gaps. Where nobody speaks, the music is the lead at −4 dB with no ducking. A level you state always wins.
- It handles the ends. Music fades in over 1 second and out over 2 by default, and a track shorter than the section it covers loops to fill it, so nothing stops dead mid-bar.
- It revises in place. Retime, refade, loop or change the ducking without losing the other settings, swap in a different song at the same position and level, or move your cuts onto the music with beat sync.
Hear Smooth Ducking
A 12-second test: a steady synthesized chord plays as the music, and a higher tone stands in for a voice at 0:02–0:04 and 0:06–0:08. With smooth ducking on (B), Valmera’s renderer pulls the chord down about 14 dB while the tone plays and restores it afterward.
Download the measured results (JSON) or the steps to reproduce them (README).
Settings That Matter
| Setting | Default | How to change it |
|---|---|---|
| Where it plays | The whole video | “Start the music at 0:12 and stop before the outro.” |
| Where the song starts | The top of the song | “Skip the intro.” “Start at the chorus.” |
| Level under speech | −18 dB bed, smooth ducking | “Quieter under my voice.” “Don’t duck it.” |
| Level with no speech | −4 dB lead, no ducking | “Louder in the montage.” |
| Fades | 1 s in, 2 s out | “Fade out over the last 4 seconds.” |
| Short tracks | Loop to fill the video | “Don’t loop it; let the music end with the song.” |
Where to Get Music You Can Use
Valmera has no stock music library, so the track, and the right to use it, come from you:
- Your own or licensed music. Tracks you made, commissioned or licensed. Upload the file, paste a link (YouTube, SoundCloud, Dropbox, Google Drive), or upload a video and Valmera uses only its audio.
- A song you name. Say “add [song] by [artist]” and Valmera finds the official recording (not a cover or sped-up edit) and tells you which version it used. That confirms the recording, not a license: a commercial song needs permission, or the video can be claimed, muted or taken down.
- Trending TikTok and Reels sounds. These are licensed inside the apps. Have Valmera cut your edit to the sound’s beat, export a version without it, then add the official sound in the app.
- Royalty-free and Creative Commons tracks. Check the license: some require credit, some forbid commercial use. The Creative Commons FAQ covers CC music in video.
Remove the Music Already in Your Footage
When the recording already has music baked in (a vlog shot in a café, a song under the talking), Valmera splits the original soundtrack into speech and everything else, so you can mute the music and keep the voice, duck it 12 dB, or keep only the music. The split is reversible but not surgical: on a dense mix a trace of the muted side can remain, so listen to the preview.
Limits
Valmera fits music; it doesn’t make it. There is no AI music generation and no built-in stock music library, so you bring the track and the license for it. There is no audio denoise, so background noise in a recording stays under the music. Music separation works on the footage’s own soundtrack and has a maximum video length (the agent tells you if yours is longer).
Part of a Full Mix
Move your cuts onto the music with beat sync, add sound effects where a moment needs a hit, and set the final loudness with the audio loudness targets. The sidechain ducking entry explains the dips, the audio and music docs cover every parameter, and the YouTubers page shows music in a complete edit.
Using Claude, ChatGPT or Cursor? The same add_music, set_music_fit and get_audio_analysis tools are available through Valmera’s MCP server.
Frequently Asked Questions
Made with Valmera: real exports
Fit Music to Your Next Video
Bring the track, describe the feel, review the preview and export a full-quality MP4. Creating an account and uploading are free; editing requires a subscription.
Create an account →