Demucs: AI Audio Editor for Content Creators
Which audio editor splits a song into vocal and instrumental stems without installing Python or CUDA? Demucs, a Audio Editor tool, runs the open-source Demucs v4 model in the browser for content creators making karaoke tracks. Audio researchers needing local batch processing at scale should consider the command-line Demucs project instead.
What Demucs Does
For content creators and producers, Demucs separates a single mixed audio file into isolated stems without requiring any local AI setup.
The hosted version at Demucs.net runs the upstream facebookresearch/demucs model in the cloud, so users upload a track, pick a configuration, and download results instead of configuring PyTorch or a GPU. It offers 2-stem splits for vocal removal, 4-stem splits for vocals, drums, bass and other instruments, and a 6-stem option adding guitar and piano. One credit covers about one minute of audio on the standard model.
Main Features
Music Source Separation
Demucs splits an uploaded track into vocals, drums, bass and other instruments using the Demucs v4 model. The hybrid architecture processes both waveform and spectrogram data, which is why separated stems carry fewer artifacts than older separators. This lets creators pull a clean instrumental or acapella from a single upload rather than re-recording parts.
Multiple Model Support
Three Demucs v4 variants are available: htdemucs for speed, htdemucs_ft for best quality, and htdemucs_6s for six stems including guitar and piano. Each model costs a different number of credits per job, so users trade processing time and credit cost against output precision depending on the task at hand.
Multiple Output Formats
Stems export as 320 kbps MP3 on every plan, with 16-bit WAV or 24-bit FLAC added on the Studio plan. Lossless files cost the same credits as MP3, but run four to five times larger, so a slow connection means longer preview load times for WAV or FLAC downloads.
Simultaneous Separations
Pro accounts run two separations at once and Studio accounts run three, compared to one on the free plan. This lets a Pro or Studio user queue a batch of tracks together instead of waiting for each file to finish before starting the next one.
Use Cases
-
Create Karaoke and Instrumental Tracks
Content creators upload a song and use Demucs’s 2-stem separation to produce a vocal-free instrumental for video or personal use. The built-in stem player confirms quality before export, letting creators check results before spending credits on a higher-quality model.
-
Isolate Vocals for Remixes
DJs and music producers separate vocals from a track with Demucs to build an acapella for mashups or live sets. The htdemucs_ft model gives cleaner isolation for performance use, where leftover vocal bleed would be unwanted.
-
Transcribe Drum and Bass Parts
Musicians isolate drum and bass stems with Demucs’s 4-stem or 6-stem configuration to study patterns from a reference track. Soloing a stem in the player helps transcribe a part without the full mix interfering.
Best For / Not For
Demucs is built for people who need a quick stem split without running the model locally.
Content Creators producing karaoke or backing tracks, DJs and Music Producers isolating vocals for remixes, and Musicians transcribing drum and bass parts are the core audiences the site is built around.
Demucs’s free tier caps files at 6 minutes and 50 MB with a single separation at a time, which rules it out for anyone processing long sets or full albums without a paid plan or credit pack.
Pricing
Demucs.net offers a free plan, a Pro plan at $9 / user/month, billed monthly, a Studio plan at $19 / user/month, billed monthly, and one-time credit packs from $5 / one-time to $49 / one-time.
| Plan | Price | Included |
|---|---|---|
| Free | Free | 15 credits per month, 3 four-stem songs or 5 vocal removals per month, 6-minute file limit, 50 MB max file size, stems kept for 24 hours |
| Credit Pack 120 | $5 / one-time | 120 credits valid for 365 days, approximately 30 songs, Pro-level file limits |
| Pro | $9 / user/month, billed monthly | 400 credits per month, approximately 100 four-stem songs per month, 15-minute file limit, 200 MB max file size, 2 simultaneous separations, stems kept for 7 days |
| Studio | $19 / user/month, billed monthly | 900 credits per month, approximately 225 four-stem songs per month, 15-minute file limit, 200 MB max file size, 3 simultaneous separations, stems kept for 30 days |
| Credit Pack 600 | $19 / one-time | 600 credits valid for 365 days, approximately 150 songs, Pro-level file limits |
| Credit Pack 2000 | $49 / one-time | 2,000 credits valid for 365 days, approximately 500 songs, Pro-level file limits |
Free and Pro credits expire 30 days after they arrive, while credits bought in a pack last 365 days instead.
Pricing checked 2026-10-04.
Quick Comparison
AudioStrip is Demucs’s main alternative for quick vocal and instrumental extraction from streaming links. AudioStrip pulls audio directly from YouTube and Spotify URLs without a manual upload step. Demucs counters with three selectable Demucs v4 models, including a 6-stem option that adds guitar and piano. Choose AudioStrip if the source is a streaming link rather than a local file. Choose Demucs if model choice and lossless WAV or FLAC export matter more.
Verdict
Demucs separates tracks into up to 6 stems, with Pro offering 400 credits a month at $9 / user/month for about 100 songs. It fits creators and producers doing occasional stem work, not teams needing an API or third-party integrations.
