Separates the noise from a stem.
Separates the noise mixed into vocal and instrument stems. It works best on noisy stems, such as AI-generated or AI-separated ones. Clean before you mix and the noise won't rise with the sound when you boost it with EQ, and the compressor reacts less to the noise. Pick the model that matches your source. Drums are not supported.
Sign in to get started.
Sign inCredits
14credits / 1 min
≈ $0.14 · incl. tax
3 min 42 credits · 5 min 70 credits
What you can upload
Length~5:00min
ChannelsMono / Stereo
Bit depth~32bit
Sample rate~48kHz
The output is 24bit / 44.1kHz.
Processing time · About 30 seconds for 5 minutes of audio.
Results kept for 24h
24 hours after a job finishes the files are deleted from the server and cannot be downloaded again. Please download your results right away.
Unsupported sources
Pick the model that matches your source. Drums are not supported.
Models used
DECANT by letheaudio
in-house
Fuzzy STN Decomposition
implemented from Fierro & Välimäki, JAES 2023