Demucs Audio Separator (4/6 stems) — ComfyUI node from Audio Separation (Demix). Run it in the cloud, no install required.

Audio separation (aka demixing) nodes, for Vocals, Instruments, Bass, Drums and Others (experimental Piano and Guitar).
Using MDX-Net and Demucs, no extra dependencies, support for batch and resample.
Choose between High Quality and Speed. All safetensor models (No ONNX, No PyTorch)

Audio Separation (Demix)

input_sound

model

shifts

overlap

custom_segment

segment

target_device

AudioSeparateDemucs

vocals

drums

bass

other

guitar

piano

Demucs Audio Separator

Name	Type	Default	Description
input_sound	AUDIO		—
model	COMBO		13 options: ⬇️ Hybrid Transformer fine-tuned, ⬇️ Hybrid MDX 1st 2021A, ⬇️ Hybrid MDX 2nd 2021B, ⬇️ Hybrid MDX A Rep. HO, ⬇️ Hybrid MDX A Rep. TO, ⬇️ Hybrid MDX A Reprod., +7
shifts	INT	00–16	Number of random shifts for equivariant stabilization. Higher values improve quality but are slower. 0 disables it.
overlap	FLOAT	0.250–0.99	Amount of overlap between audio chunks. Higher values can reduce stitching artifacts but are slower.
custom_segment	BOOLEAN	false	Enable to override the model's default segment length. Disabling uses the recommended length from the model file. Useful for HDemucs and Demucs models, not much for HTDemucs.
segment	INT	4410–120	Length of audio chunks to process at a time (in seconds). Higher values need more VRAM but can improve quality.
target_device	COMBO	cpu	The device (CPU or CUDA) to which the projection layer will be assigned for computation.

Demucs Audio Separator

Inputs (7)

Outputs (6)