Run Demucs
Ready in about 90 seconds. No install, no setup. From $0.45 per hour.
What it does
Demucs takes a mixed audio file and separates it into individual tracks. Vocals come out on one stem, drums on another, bass on a third, and the rest on a fourth.
For video work the useful trick is pulling the dialogue away from the music and background noise on a recording where they were never separate.
For music work it is the standard tool for making instrumentals, isolating a vocal for a remix, or rescuing a recording where one element was mixed too loud.
What you can make
These are prompts, not fixed presets. The workspace opens with one of them already loaded so your first result is one click away.
Your setup
We run this on an RTX 4090. It separates about 8 minutes of audio per minute of GPU time. $0.45 per hour, billed by the minute.
Why this card?
Demucs works on the raw waveform rather than a compressed representation, which means it moves a great deal of data for every second of audio. That makes memory bandwidth the limit, not model size. The RTX 4090 processes a track roughly two and a half times faster than an RTX 3090 for twice the price, so it is both quicker and slightly cheaper per track.
- Memory
- 24GB
- Speed
- about 8 seconds per minute of audio
- Rate
- $0.45/hr
What will it cost you?
Billed by the minute on a RTX 4090 at $0.45 per hour. A batch job shuts the GPU off when the last minute of audio is done, so this is the whole cost.
Two ways to run it
Click and use it
Upload a track, listen to each stem, download the ones you want.
Get early access Best for exploring.Hand us the whole pile
Process a whole library into stems in one run.
Join the list Best for volume. The GPU shuts off automatically when it is done.Model facts
| Parameters | 80M |
|---|---|
| License | MIT |
| Memory required | 8GB minimum |
| Base model | Hybrid Transformer Demucs v4 |
| Publisher | Meta |
| Recommended hardware | RTX 4090, 24GB |
| Hugging Face | facebook/demucs |
Want Demucs the day it opens?
We are testing the rental flow with a small group first. Leave an email and we will tell you when Demucs is ready to run. One message, no newsletter.
Demucs questions
Do I need to install anything to use Demucs?
No. We start a machine with Demucs already loaded and hand you a link. Everything runs in your browser, there is nothing to download, and nothing is left on your computer afterwards. It works the same on a Mac, a Windows laptop, or a Chromebook. A workspace is usually ready in about 90 seconds.
How much does it cost to run 120 minutes of audio?
About $0.11. Demucs takes about 8 seconds per minute of audio on the RTX 4090 we recommend, so 120 minutes of audio is roughly 15 minutes of GPU time at $0.45 per hour. Billing is by the minute, and a batch job shuts the GPU off the moment the last item finishes. The calculator above works this out for your own numbers.
Can I run Demucs on a cheaper card?
Sometimes, and the calculator will not always make it look worth it. Demucs works on the raw waveform rather than a compressed representation, which means it moves a great deal of data for every second of audio. That makes memory bandwidth the limit, not model size. The RTX 4090 processes a track roughly two and a half times faster than an RTX 3090 for twice the price, so it is both quicker and slightly cheaper per track. If you want to try a different card anyway, the advanced catalog lets you pick one and shows the estimated time before you commit.
Can I use Demucs commercially?
Yes. Demucs is released under the MIT license, which permits commercial use with no conditions worth worrying about. You can use the output in paid client work and in products you sell.
What happens to my files after I stop?
Anything you save into your workspace folder stays in your account and is there when you come back. Everything else is destroyed: the machine is terminated and its disk is wiped. Batch job inputs and outputs are deleted from our storage 24 hours after you download them. We do not read your files or use them to train anything.
What if it fails partway through?
You are not charged for work that did not complete, and failed sessions are refunded automatically. Batch jobs write each output as it finishes rather than at the end, so a failure at item 380 of 400 leaves you 379 usable files and a message saying what broke. Restarting picks up where it stopped rather than redoing the work.
Own a GPU that sits idle? Put it to work. The average provider earns $180 to $420 per month per card.
List your GPU