RemoveBackgroundMusicBuilt for finished files
Use an MP4, MP3 or WAV even when you no longer have the editor timeline or isolated music stem.
VOICE FORWARD · MUSIC OUT
Remove music from video and keep the voice.
Bring interviews, narration and conversation forward without reopening the original edit. The video stays on your device; only extracted audio is processed.
Restoring your workspace…
Dialogue only is selected for new files. It focuses on speech and may remove ambience and effects; choose Dialogue & sound effects when those sounds matter.
- Spoken voice stays in sync
- Original video never uploads
- Video or WAV export, depending on source
QUICK ANSWER
What this workflow actually does.
This page starts in Dialogue only mode: the service analyses the extracted soundtrack to isolate people speaking while reducing music and other sounds. Choose Dialogue & sound effects if you also need scene sounds. Neither mode creates a transcript or a synthetic voice. The result is usually strongest when speech is distinct from the soundtrack; singing, heavy mastering and instruments in the speech range can cause residue or missing detail.
Hear the same moment
The player places the original beside the processed preview so you can judge intelligibility and artifacts directly.
Choose one-time processing
Choose Single File for one eligible short file, or use a prepaid minute balance across multiple clips. Neither option is a subscription.
Preparing captions or notes? Check whether cleanup improves transcription. Quieter music does not guarantee that the words were preserved.
USE IT WHEN
The edit is finished, but the soundtrack is not.
The browser keeps the original MP4 local, extracts only its audio, and lets you compare up to 15 seconds free before choosing single-file processing, a minute pack or a subscription.
- Recover narration under an over-loud soundtrack
- Create cleaner clips for captions and transcription
- Prepare commentary for reuse on another platform
- Make dialogue easier to follow before adding new music
A SAFER WORKFLOW
Test the difficult moment before processing the full file.
- 01
Use the cleanest authorised file
Choose the original editor export, camera file or WAV where possible. Avoid a screen recording or a copy that has already been denoised, compressed or separated.
- 02
Find the hardest spoken passage
Preview quiet speech under loud music, names and numbers, consonants, breaths and a transition. An easy sentence is not enough to approve the full recording.
- 03
Compare intelligibility and timbre
Match the apparent voice level before switching. Check that every word remains understandable and that the speaker does not become metallic, watery or inconsistent.
- 04
Choose what to keep
Choose Dialogue only to focus on people speaking, or Dialogue & sound effects when the surrounding scene sounds matter. A new mode needs its own preview.
ILLUSTRATIVE BEFORE / AFTER
Compressed speech and music
The source passed through a 48 kbps MP3 encode/decode before separation. Compare consonants, breaths and pauses; separation cannot restore detail already lost to compression.
Project-owned synthetic test material. Existing project-owned synthetic voice-and-music mix, procedural additions and a low-bitrate round-trip. This 15-second example is not a customer testimonial or a promise about your file.
Sample limit: This shares a parent voice with the earlier easy-mix example. It does not validate every codec, singing voice or long podcast. Both players use delivery MP3 encoding. Match the apparent voice level when comparing; this is not a normalized loudness benchmark.
LISTENING DECISION
How to judge the preview.
For voice-led material, intelligibility and a stable speaker identity matter more than absolute silence between words.
| What you hear | What it means | What to do |
|---|---|---|
| Every word remains clear | Speech detail survived the separation. | Continue, then check the entire full-length result for local failures. |
| Voice sounds thin or underwater | Part of the speech was estimated as music. | Compare the other mode with an available preview credit, or return to a cleaner source. Do not repeatedly process the same damaged output. |
| Consonants or names disappear | Important high-frequency speech overlaps the soundtrack. | Do not approve the full result; test another section or use lighter removal. |
| Breaths and room tone cut in and out | The residual estimate is unstable around low-level sound. | Keep a little original ambience or repair those regions in an editor. |
PREVIEW CHECKLIST
Use a revealing 15-second excerpt.
- Use quiet and loud speech in the same preview.
- Include names, numbers and sentence endings.
- Compare at matched voice loudness.
- Check pauses, breaths and room tone.
- Listen once on a phone speaker.
IMPORTANT LIMITS
Know when to stop.
- Speech and singing can be ambiguous to a separator.
- Dense music in the vocal range can remove consonant detail.
- A second pass usually magnifies artifacts.
- For broadcast or legal material, keep the original and document every change.
EVIDENCE AND LIMITS
Why the preview matters.
Published research describes real-world soundtrack separation as a three-stem problem involving speech, music and sound effects. It also reports that separation can introduce artifacts and that remixing some estimated interference can improve downstream results. Those findings do not predict the quality of any individual upload. Choose Dialogue & sound effects to retain scene sounds, or Dialogue only to focus on people speaking; compare your own preview before paying.
- The Cocktail Fork Problem: Three-Stem Audio Separation for Real-World Soundtracks
- Tackling the Cocktail Fork Problem for Separation and Transcription of Real-World Soundtracks
These are independent research sources, not endorsements of RemoveBackgroundMusic. Reviewed . See the editorial and testing policy for how technical claims and product observations are separated.
FOR YOUR REGULAR WORKFLOW
Monthly minutes. Your choice of billing.
One account for music removal, vocal removal and replacement.
Creator
$16.58/mo
$199 billed annually
2 months freeSave $40.88 vs monthly
For regular personal projects.
30 minutes added each month
- Use across supported audio tools
- Unused minutes expire each service month
- 30-second minimum per new AI job
- Cancel renewal; paid service continues through your year
ProRecommended
$24.92/mo
$299 billed annually
2 months freeSave $60.88 vs monthly
For ongoing video production.
60 minutes added each month
- Use across supported audio tools
- Unused minutes expire each service month
- 30-second minimum per new AI job
- Cancel renewal; paid service continues through your year
Studio
$45.75/mo
$549 billed annually
2 months freeSave $110.88 vs monthly
For higher-volume production.
120 minutes added each month
- Use across supported audio tools
- Unused minutes expire each service month
- 30-second minimum per new AI job
- Cancel renewal; paid service continues through your year
Paid yearly. Added monthly. Yearly plans charge the full annual price now and release minutes monthly, not all at once.
No rollover. Unused subscription minutes do not roll over.
Cancel renewal anytime. Already-paid service continues through your paid period.
Prices in USD; applicable taxes shown at checkout. “2 months free” compares annual pricing with 12 monthly payments; no extra months are added. Billing & refund FAQs
FOR OCCASIONAL PROJECTS
Prefer a one-time pack?
Single-file prices are shown in the tool after your preview.
Quick Fix
$8.9910 minutes
For occasional projects and a few clips.
- One-time payment · No automatic renewal
- Shared minutes for previews and full files
- 30-second minimum per new AI job
- Minutes expire 12 months after purchase
Creator Pack
$34.9950 minutes
For a larger collection of files, at your own pace.
- One-time payment · No automatic renewal
- Shared minutes for previews and full files
- 30-second minimum per new AI job
- Minutes expire 12 months after purchase
QUESTIONS FOR THIS WORKFLOW
Before you process the full file.
Compare subscriptions, one-time purchases and charging rules on the Pricing FAQ.
Is this the same as a vocal remover?
No. A vocal remover usually targets singing in a music mix. This workflow is intended to keep spoken dialogue and non-music context while reducing the background score.
Can it keep two speakers?
It can keep multiple speakers, but overlapping speech, music and singing can make the estimate harder. Test a passage containing both speakers.
Will the voice be regenerated?
No. The service separates the existing soundtrack; it does not clone, transcribe or synthesize the speaker.
Can cleaner dialogue improve captions or transcription?
It may help when music masks words, but separation can also remove consonants or change speech. Compare original and processed audio with the same recognition settings, and check names, numbers and negations against the recording. Compare actual transcription errors.
Will spoken dialogue survive when background singing overlaps it?
It can try to keep speaking while reducing music, but the voices can overlap in pitch and timing. Singing may remain and quiet words may be damaged. Preview both voices together and check consonants, names and voice texture before full processing. Choose the voice you need.
Should I choose Dialogue only or Dialogue & sound effects?
Choose Dialogue only when intelligible speech is the priority. Choose Dialogue & sound effects when room tone, movement or scene sounds matter. Neither guarantees perfect recovery; another mode needs an available preview allowance or a confirmed balance preview. Compare the scene-sound workflow.
Want to remove a voice instead of the music? Try the video vocal remover. Choose singing or dialogue, then compare a preview; other voices and sounds can also be affected.
YOUR FILE IS THE TEST
Hear what the model keeps before you pay.
Results vary when voice, music and effects overlap. Start with the preview and compare the same moment. Continue with Single File for an eligible short file, or use minutes from a pack or subscription, only if the result works for your source.
Choose a video or audio file