Press "Enter" to skip to content

KageMusic (Shadow Music)

KageMusic is a Music Source Separation based on Deezer Spleeter, Google TensorFlow, allow you to split or extract Voice and Music or Multiple Instrument into different files or tracks.

KageMusic also can be use to make a A cappella (Vocal) or Karaoke (Instrumental)

KageMusic also can separate drums / bass / piano / other separation


User Interface

With more Options & Queue Support


Here several example processed audio file

Super Idol (Bahasa Malaysia Cover) feat. Liliana Vampaia

Instrumental (VR)
Vocal (A.I. Facebook Research)

Kimi Ga Inai Natsu

Original Source
A capella
Instrument (Karaoke)

fripSide – Prominence

Original Source
A capella
Instrument (Karaoke)

Detective Conan Opening – Episode 86

A capella
Instrument (Karaoke)

Detective Conan Ending – Episode 86

A capella
Instrument (Karaoke)

As you can hear, not all music can be process properly due to lossy codec nature, it’s advisable to use Hi-Fi Song such as FLAC codec in 48KHz or at least CD level quality (44.1KHz @ 16-bit).

MP3, MP4 might not produce good enough, use high bit-rate is top priority to use this program.

In the future, A.I. smart enough to reconstruct missing information from lossy codec and produce higher quality results.

System Requirement

ProcessorIntel Core i3 4th Gen / AMD FX-4100 (AVX & LibROSA)
Memory8 GB
Graphics CardNvidia RTX 2000 series / AMD Radeon RX 5000 series (TensorFlow)
Storage3 GB Free Space for A.I. Engine and Training Data
Operating SystemWindows 7 SP1 64-bit with .NET Framework 4.8

If you have weak (Intel Atom, Celeron, Pentium) or old (Core 2 Duo, AthlonII) might not able to run properly, or might overheat because it run forever, since these CPU have missing modern Instruction Set such as AVX-128

This program can be run without GPU, since it run on CPU, processing time might take longer on 4 core system.


HITOHA Cloud (Malaysia)
SourceForge (Global)
GitHub Release (Global)
Currently have one active link