Ijharul Islam
To dub a video and keep the background music, you need the music and sound effects without the original voice. Use the separate music track from the original edit if you have it; if you only have the finished video, an AI source-separation model such as Demucs splits the voice from the music, and you lay the new voice over the music-and-effects track.
If neither is possible, the fallback is voice-over style: lower the whole original track and play the new voice on top. This guide explains all three methods, how to do them for free or with AI, and how to fix common problems.
The short answer
Keep the background music by replacing only the voice. Best: mix the new voice over the original music and effects stems from the edit. Next best: run the finished video through an AI source-separation model (like Demucs) to remove the voice, then mix the dub over what's left. Fallback: lower the original audio and play the dub on top, voice-over style.
Key takeaways
- The original project's separate music and effects tracks give the cleanest result.
- AI separation recovers the music from a finished video, with small artefacts.
- Voice-over style keeps everything but leaves the original voice faintly audible.
- Keep music under speech quiet so the new voice stays clear.
Disclosure: we make QuickVeed and show how to do this in it, along with other ways. Details about other products come from their public pages, checked in October 2026 and linked at the end.
Three ways to dub a video and keep the background music
| Method | Quality | What you need | Best for |
|---|---|---|---|
| Original stems (M&E track) | Best | The editor's separate voice, music and effects tracks | Your own productions |
| AI voice separation | Very good, small artefacts | Only the finished video | Old videos, videos from others with permission |
| Voice-over style (ducking) | Original voice faintly audible | Only the finished video | Interviews, news-style, quick versions |
Film studios deliver a music-and-effects (M&E) track with every film for exactly this reason: dubbing studios replace the dialogue and keep the rest. If you edit your own videos, export a version without the voice track whenever you finish a project. It costs nothing and makes every future dub cleaner.
How AI voice separation works
Source-separation models are trained on music where each instrument and voice is recorded separately, so they learn to pull a mixed track apart. Meta's open-source Demucs, for example, splits audio into vocals, drums, bass and other sounds. For dubbing, you keep everything except the vocals as the new background, and the vocals give a clean sample of the speaker if you want to clone their voice.
- What works well: speech over background music, room tone, most sound effects.
- What's harder: songs with lyrics (the singing is removed with the voice), crowd noise, heavy reverb, and speech recorded through the same microphone as loud music.
- Typical artefacts: a slight watery sound in the music where the voice used to be, or faint traces of the original words.
You can run Demucs yourself for free if you're comfortable with the command line; it's on GitHub. Many audio apps and AI dubbing tools build it in.
Manual method: separate, then mix
- Extract the audio from the video. Our free extract audio from video tool does this in the browser.
- Run the audio through a separation model to get a voice track and a music-and-effects track.
- Record or generate the dub, one clip per line, in the new language.
- In an audio or video editor, place the music-and-effects track under the video and put each dubbed line where the original line started.
- Set the voice so it sits clearly above the music. Lower the music slightly under speech if it competes.
- Export and listen on phone speakers and headphones before you publish.
Keep the timing of lines close to the original. If a translated line runs into the next one, shorten the translation rather than speeding the voice up a lot. Our free audio waveform generator helps you see where speech and pauses fall.
Keep the music when dubbing in QuickVeed
In QuickVeed AI dubbing this is a choice you make before preparing a video (up to 10 minutes):
| Option | What it does | Credits |
|---|---|---|
| Keep music and effects | AI (Demucs) separates the voice from the music and effects; the dub is mixed over the music and effects | 9 per minute |
| Voice-over style | No separation; the original audio is lowered under the new voice | 1.5 per minute, rounded up |
- Upload the video or pick it from your library.
- Choose Keep music and effects and click Prepare for dubbing.
- Fix the transcript and save it.
- Under Dub it, pick the speaker's cloned voice (30 credits a minute per language) or a stock voice (5), choose the language and click Dub.
- Download the MP4 with the new voice over your original music, plus SRT and VTT files.
Preparing happens once per video, so the separated music is reused for every language you dub. A 7-minute video costs 63 credits to prepare and 210 per language in the speaker's voice. The original video picture is kept as it is. Only clone a voice you have the right to use: your own, or a speaker who has agreed to it in writing. QuickVeed requires that consent, and so should any tool you use.
Mixing levels that keep the voice clear
Once the music is back under the new voice, the balance between them decides whether the dub sounds professional. Some practical rules:
- Voice first. Speech should be clearly louder than the music whenever someone talks. If you have to strain to hear a word on phone speakers, the music is too loud.
- Duck, don't mute. Lower the music a little under speech and let it come back up in pauses, so the video still feels alive.
- Match the original. Listen to a part of the original where the speaker talks over music, and match how loud the music felt there.
- Keep sound effects. Clicks, whooshes and room sound make a dub feel real; separation keeps them in the effects track.
- Check the ending. Outros often have louder music; make sure the last line of speech isn't buried.
- Listen on three devices: phone speaker, headphones and a laptop. Most viewers are on a phone.
AI dubbing tools set these levels for you. If a part sounds off, it is usually a line that runs into loud music; shortening that line so it lands in a quieter moment fixes more than any volume change.
Fixing common problems
| Problem | Fix |
|---|---|
| Faint original words under the dub | Use separation rather than voice-over style, or lower the music track a little more under speech |
| Song lyrics disappeared | Separation treats singing as voice; use the original music file if you have it |
| Music sounds watery in places | Normal separation artefact; usually masked by the new voice |
| New voice too quiet on phones | Raise the voice or lower the music; check on phone speakers |
| Dub drifts out of time | Shorten the translated lines that run long |
For more on the full workflow, read how to dub a video into another language, and for YouTube uploads, multi-language audio tracks on YouTube.
Frequently asked questions
How do I remove the voice from a video but keep the music?
Use an AI source-separation model such as Demucs, which splits audio into vocals and the rest. Keep the non-vocal track as your background.
Can AI dubbing keep the original background music?
Yes, if the tool separates the voice first. QuickVeed does this with its Keep music and effects option for 9 credits a minute.
What is an M&E track?
A music-and-effects track: the full soundtrack of a video without dialogue. Studios make it so films can be dubbed without losing the music.
Why can I still hear the original voice in my dub?
The original audio was lowered rather than separated (voice-over style), or the separation left traces. Use separation, or lower the background more under speech.
Will separating the voice affect copyright?
Separation doesn't change who owns the music. You need the same rights to use it in the dub as in the original video.
Try it with your own video
Dub one video with its music kept and compare it with the original. Upload a short video, dub it into one language, and judge the result for yourself. QuickVeed plans start at $18/month, and new accounts get a 14-day trial with 75 credits, no card needed. Sign up or see pricing.
Sources
Competitor details in this post were checked against these public pages in October 2026. Products and prices change, so confirm with each vendor before you decide.
