Remove background music but keep the voice: choose the right task
“Remove music” can describe two very different jobs: extracting a sung vocal from a song, or preserving spoken dialogue under a soundtrack. Start by identifying which voice you need. A music vocal model is a reasonable starting point for singing; a dialogue-focused processor is designed for speech and background sound. Their names can look similar while the intended targets differ.
1. One-minute audio test
So Much is Wrong — gurdonark, featuring SackJo22 · 1:57–2:57 · 60s
This is an existing spoken-word/music mix by the credited artist, not a synthetic voice-over example. Compare speech loss and remaining background music.
Click or drag a waveform to seek. On a focused waveform: ←/↓ back, →/↑ forward; Tab next track, Shift+Tab previous track. Space plays or pauses.
Loop
Test settings and processing record
Tested on 7 November 2026: standard mode, 4-second segments, NVIDIA RTX 4060. Times cover local inference and export, excluding model loading, queue and transfers.
| Model | Processing time | Peak VRAM |
|---|---|---|
| kim-melbandroformer | 7.45 s | 1,449.2 MiB |
Model loading, queue and transfer time are excluded. All versions use the same input and standard mode. MP3 previews: 192 kbps; references: 96 kbps mono, 16 kHz.
Author reference audio — not AI output
Reference recordings were aligned to the original mix to check that the target sound occurs in this passage. This verifies activity, not subjective separation quality. Reference levels were adjusted; these are not the original mix balance.
Full test record ↓2. Identify the recording and the source files you have
- A singer over instruments: try vocal/instrumental separation and retain Vocals.
- A presenter speaking over a music bed: look for a speech or dialogue isolation tool, and test actual word intelligibility.
- A voice with hiss, hum or traffic but no music: investigate noise cleanup rather than assuming a stem splitter is the right repair.
- Two people talking over each other: this is a multiple-speaker problem; removing music alone will not assign each speaker a separate track.
uvr online is primarily a music-separation tool. Its standard vocal output is not a promise of professional dialogue restoration or named-speaker extraction. For a speech-only recording, treat a vocal-model trial as an experiment, not a guaranteed fix. This guide explains how to decide when to stop and use a more suitable workflow.
If you created the video or podcast and still have its project, return to the separate microphone and music tracks. Lowering or muting the original music track avoids asking a model to undo a finished mix. If you only have a video, export an authorized audio file in a video editor first; this website’s upload workflow accepts audio files, not arbitrary video links. Keep the unprocessed recording so you can restore a damaged word in an editor if appropriate.
3. Check for missing words and remaining music
After the full result is ready, compare the difficult passage with the original. Check both the wanted detail and unwanted leakage before choosing a track.
4. Use this symptom-based checklist
Online models cover vocals/instrumental, four or six stems, instrument extraction, lead/backing vocals and dereverb. Check availability in the selector. Outputs are estimates, not named-speaker identification, general noise repair or guaranteed recovery of every sound.
5. Frequently asked questions
Can an instrumental remover keep spoken voice?
It may retain some speech, but a singing-oriented model is not a dialogue guarantee. Test quiet words and consider a dedicated speech workflow.
Is background music removal the same as noise reduction?
No. A soundtrack, a steady hum and an isolated click present different problems. Identify the unwanted sound before selecting a processor.
Will this separate two speakers?
A standard vocal/music split groups voice-like content. It does not provide one independent file per person.
Can a background music remover remove music from audio with speech?
That is the intended goal of a speech-focused music remover. It still needs an intelligibility check; a music vocal model is not an interchangeable substitute for every dialogue recording.
Audio sources and usage
So Much is Wrong — gurdonark, featuring SackJo22 · CC BY 3.0
Includes source material from: SackJo22 / 51993
The excerpt was trimmed, separated into audio stems and encoded as MP3.
If a sample affects your privacy or rights, contact support@uvr.online or open a Support ticket with the sample link to request a review and removal. Contact support →