01
Following a UVR guide
The tutorial names a model; pick the same family here instead of installing the app and downloading model files.
UVR Online
Choose the stems to create
Get a clean vocal track and an instrumental backing. Perfect for karaoke and covers.
Detailed splits
Output format
Estimated costAbout 80 credits
UVR Online
The models from your UVR tutorial, without the install
Ultimate Vocal Remover runs open-source separation models on your own computer. Here two of those model families, BS-RoFormer and Demucs v4, run in the cloud: choose one by name, upload a song and download the stems.
How it works
Pick a model, pick the stems, download
1STEP 1
Upload the song
Sign in and add an MP3, WAV, FLAC, AAC, M4A or OGG file up to 50 MB and 12 minutes.
2STEP 2
Choose the model and stems
Select BS-RoFormer or Demucs v4 in the model menu, then vocals and instrumental, four stems or six stems.
3STEP 3
Download WAV or MP3
WAV is selected by default. Download each stem on its own or all at once, and every result is labeled with the model that made it.
What you get
Ultimate Vocal Remover online, with every model labeled by name
You choose the model yourself instead of a hidden preset, and the result tells you which one ran.
Choose the model by name
The model menu lists BS-RoFormer and Demucs v4 instead of a hidden preset. BS-RoFormer uses a band-split RoFormer vocal model for two stems and a six-stem RoFormer for four or six. Demucs v4 runs htdemucs_ft, or htdemucs_6s for six stems, and uses fewer credits per run. The estimate updates as soon as you switch.
Compare BS-RoFormer and Demucs on one song
Your history labels every separation with the model that made it, and the top of the result repeats the name. Run the same song once with BS-RoFormer and once with Demucs v4, then move between the two results and solo the vocals in each to hear which model handles your track better.
No GPU, no install
Separation runs on cloud hardware, so there is no Python, no GPU driver and no model file to download. Your computer only uploads the song and plays back the stems, which means a laptop, a Chromebook or a shared computer is enough. You can leave the page while it works; finished results stay in your history.
Two stems to six
Pick vocals and instrumental for an acapella and a karaoke track, a four-stem split with drums, bass and other, or six stems with guitar and piano on their own tracks. Both model families run all three. Optional detailed splits add lead and backing vocals, drum kit pieces or lead and rhythm guitar, for extra credits.
Mix and export in the browser
Mute, solo, set volume and pan per stem, or switch to the original recording to compare. Export remix then renders your own mix as a WAV file without opening a DAW. Stems download as WAV, which is selected by default, or as MP3, one at a time or together with Download all.
Who it's for
For anyone a vocal remover tutorial left at the install step
If a guide names BS-RoFormer or Demucs, you can run it here from the browser.
02
Computers without a capable GPU
Large separation models can take a long time on an ordinary CPU. Here the work happens on cloud hardware.
03
Covers and karaoke
Get a clean acapella and instrumental from a two-stem split, ready to sing over or remix.
04
Producers building in a DAW
Take six WAV stems into your session and rebuild, sample or rebalance the parts.
Models
Which model to pick
Two open-source model families are available today. Both can run the 2, 4 or 6-stem split.
01
BS-RoFormer online
Band-split RoFormer. Two stems use a dedicated vocal model; four and six stems come from a six-stem RoFormer model. It is the default in the model menu.
02
Demucs v4 online
Run Demucs online without installing Python or a GPU driver. Hybrid Transformer Demucs: htdemucs_ft for two and four stems, htdemucs_6s for six. It uses fewer credits per separation.
03
Not available yet
MDX-Net, VR Architecture, MelBand RoFormer and ensemble modes are not on Lacuna. If your guide depends on one of them, results here will sound different.
FAQ
UVR online questions
No. Ultimate Vocal Remover (UVR) is a free, open-source desktop app, and Lacuna is not affiliated with it. This page runs two of the open-source model families UVR users know, BS-RoFormer and Demucs v4, in the cloud so you can use them from a browser.
BS-RoFormer and Demucs v4 (htdemucs_ft, plus htdemucs_6s for six stems). MDX-Net, VR Architecture, MelBand RoFormer and ensembles are not available.
No. Separation runs in the cloud. You only need a browser to upload the song, play back the stems and download them.
Every result in your history shows the model name, and the top of the result repeats it. Run the same song once with each model to compare them side by side.
Vocals and instrumental, four stems (vocals, drums, bass, other) or six stems (adds guitar and piano). Optional detailed splits can also separate lead from backing vocals, the drum kit into pieces, or lead from rhythm guitar, for extra credits.
No. Each model runs with fixed settings. Your choices are the model, the stems and the output format.
Sign in to upload. Free accounts get monthly credits that cover a separation. BS-RoFormer uses more credits per run than Demucs v4, and the estimate next to the button shows the cost before you start. Downloads are included.
Upload MP3, WAV, FLAC, AAC, M4A or OGG files up to 50 MB and 12 minutes. Stems download as WAV, the default, or MP3.
Related tools
Pages set up for one separation job
Stem Splitter
Separate vocals, drums, bass, guitar, piano and other parts.
Vocal Remover
Get the instrumental and the acapella as two files.
Karaoke Maker
Remove the singer and keep a backing track to sing over.
Lead Vocal Remover
Take out the lead vocal and keep the backing harmonies.
Drum Remover
Mute the drums and export a drumless version of the song.
Instrument Remover
Remove your instrument for a minus-one backing track.
Run BS-RoFormer or Demucs from your browser
Use UVR online: pick a model, upload a song and download the stems.