Vocal remover (remove vocals from a song)
Remove the vocals from any song to make an instrumental for covers or karaoke practice. The source separation AI HTDemucs, released by Meta, runs inside your browser, so the song is never uploaded and there is no app to install. You get two files, the instrumental and the vocals, and you can bring the vocals back in or take them out in the player before saving.
The first run downloads an AI model (about 90 MB); later runs use the copy stored in your browser. Nothing but the model is transferred and your files are never uploaded.
Settings
How to use
- 1
Add songs
Drop audio such as MP3, WAV, M4A or FLAC, or video such as MP4 or MOV. You can process several songs at once.
- 2
Choose the output format
MP3 at 320 kbps is the default. Choose WAV or FLAC if you will edit the result. The first run downloads the AI model (about 90 MB).
- 3
Listen, then save
In the result player, mute, solo or change the volume of each part to compare, then save each file or export the mix as one file.
How the vocals are removed (vs. vocal cancellers)
Old-style vocal cancellers subtract the left and right channels to cancel whatever sits in the center, so centered bass and kick drums disappear too, and reverb on the voice is left behind. Here the AI splits the song into four parts, vocals, drums, bass and other, and adds the three non-vocal parts back together as the instrumental. The bass and drums stay; only the vocals go.
In Chrome or Edge on a computer it runs on the GPU (WebGPU); on the laptop we tested (integrated GPU) a 4-minute song took about 1 minute. On iPhone, iPad and Safari it runs on the CPU and takes longer: about 7 minutes for a 4-minute song on an iPhone 15 Pro Max. The AI model (about 90 MB) is downloaded on the first run only and reused from your device afterwards.
Frequently asked questions
Is this vocal remover safe? Is my song uploaded?
Your song is not uploaded. Separation runs in the browser on your device and the audio never leaves it, so it can't be used for AI training or anything else. The network is used only to download the AI model (about 90 MB) the first time. No sign-up is needed.
Are the vocals removed completely?
In most songs they become almost inaudible, but faint traces can remain in songs with heavy reverb or backing vocals, or with many instruments in the same range as the voice. Mute the vocals in the result player to check.
Does it work on iPhone? Do I need an app?
No app is needed: open this page in Safari or another browser. On iPhone and iPad it runs on the CPU, so a 4-minute song takes about 7 minutes. Keep the screen on and this page open while it runs; switching to another app pauses the processing.
Can I publish the instrumental I made?
Publishing or distributing audio separated from a commercial song requires permission from the rights holders. Use it for your own practice, or for songs you have permission to use or wrote yourself.
Help us improve the service
If a conversion fails, sending us the circumstances helps us track down the cause. Knowing which devices and formats fail most often, and how long processing takes, lets us revise the supported formats and fix the bugs behind them. If you don't mind, leaving this on is a real help.
What is sent
The tool you used and the settings you chose (output format, quality, etc.), your device specs (GPU, memory, core count), the input format with a rough size, how many files succeeded or failed and how long it took, and the error type and details when something fails.
What is never sent
The files themselves, file names, text, subtitles or prompts you add to a video, and IP addresses. The data goes only to this site's own server, never to a third-party service.