Six-stem separation · TelkNet

Separation & stem splitGPU workload

Six-stem separation

Users who need finer stem control for learning, remixing, or teaching

Six-stem assets with optional MIDI paths for focused rhythm, low-end, vocal, and piano work

5 credits/useAudio file inputMax 50 MBEstimated 90-180 seconds

Input

Full-band songs or dense mixes split into drums, bass, guitar, piano, vocals, and other stems

Audio formats

mp3wavflacoggm4aaac

Output

Six-stem assets with optional MIDI paths for focused rhythm, low-end, vocal, and piano work

Best for

Users who need finer stem control for learning, remixing, or teaching

Models, parameters, and sources

Six-Stem Separation splits a mix into stems and transcribes selected targets through the fixed BS-RoFormer SW + multi-instrument AMT route.

BS-RoFormer SW Fixed

BS-RoFormer SW Fixed is the six-stem separator path used for vocals, bass, drums, guitar, piano, and other.

MVSep BS Roformer SW bs-roformer-infer registry python-audio-separator BS-RoFormer

Stem targets: vocals, drums, bass, guitar, piano, and other are separated before optional MIDI transcription.
Runtime asset: TelkNet prepares BS-Rofo-SW-Fixed.ckpt and config_bs_roformer_sw.yaml as local separator assets.
Public context: MVSep lists BS Roformer SW with a six-stem SDR quality table; BS-RoFormer papers document strong MUSDB18HQ and SDX23 results.

Model comparison

Model

Role

Why it is used

Output impact

ModelBS-RoFormer SW Fixed

MVSep BS Roformer SW bs-roformer-infer registry python-audio-separator BS-RoFormer

RolePrimary six-stem audio separator

Why it is usedSelected because it directly targets the six requested stems and has public MVSep/BS-RoFormer evidence.

Output impactSix WAV stems

ModelHT Demucs

facebookresearch/demucs HT Demucs paper

RoleSeparation baseline / fallback

Why it is usedHT Demucs is a strong open-source reference, but the current six-stem route uses BS-RoFormer SW.

Output impactContext or fallback stems

Paper and benchmark notes

Music Source Separation with Band-Split RoPE Transformer

SDX23 music separation track - 2023

BS-RoFormer uses band-split spectrogram features plus RoPE Transformers; the SDX23 system ranked first in the music source separation track and a smaller version reported 9.80 dB average SDR on MUSDB18HQ.

arXiv:2309.02612

Hybrid Transformers for Music Source Separation

ICASSP - 2023

HT Demucs replaces core Demucs layers with cross-domain Transformer attention and reports 9.20 dB SDR with extra data and fine-tuning.

arXiv:2211.08553

Official / repositories

MVSep BS Roformer SW bs-roformer-infer registry python-audio-separator

Papers / technical notes

BS-RoFormer

Parameter guide

Six-stem targets: choose bass, drums, guitar, piano, vocals, or other stems to transcribe.
Piano stem path: six-stem uses BS-RoFormer SW separation plus multi-instrument MIDI splitting; it does not call Aria-AMT.
Vocal harmony split: attempts to split lead vocals and harmonies.
Save separated tracks: keeps stems or track outputs for separate download.
Post-processing: quantization, dedupe, velocity smoothing, and polyphony limits all affect cleanliness.

Use cases

Practice with isolated drums, bass, guitar, piano, or vocals
Prepare stems for remixing, re-arranging, or sampling
Inspect the instrument roles inside a dense arrangement

Workflow

1Upload a song with a fairly complete mix
2Wait for the multi-stem output
3Pick the stems you want to keep working with

Pre-flight checklist

Best for richer arrangements — short clips are not recommended
More complex tracks take longer; confirm your credit budget before submitting
If you only need vocals, the dedicated vocal-separation tool is enough

Login required to use this tool

Browsing tool info is always free - no account needed.