Audio Source Separation
RASES VCVoice Canceler
Vocals and BGM, split by
three models.
RASES — Voice Canceler
What is RASES VC
Vocals and BGM,
as fast as you need.
RASES VC splits vocals from background music (BGM).
Pick Lite, Compact or Nano by quality, speed and size.
- 3modelsLite · Compact · Nano
- 8.0msFastest — Nano, NPU 100%
- 14.70dBBGM quality (SDR) — Lite
How it works
From a mix to vocals and BGM.
It lifts the voice out of a song.
Remove the vocals or use them alone — the BGM stays intact.
Models
Quality or speed.
All three split vocals and BGM.
Choose by your device and the quality you need.
- QualityRASES VCLiteQuality-first model
- BalancedRASES VCCompactBalance of quality and compute
- LightRASES VCNanoLow-compute, lightweight model
| Item | VC Lite | VC Compact | VC Nano |
|---|---|---|---|
| Time (NPU 100%) | 26.1ms | 13.9ms | 8.0ms |
| Time (NPU 25%) | 104.3ms | 55.8ms | 32.2ms |
| Library size | 22.21MB | 17.83MB | 10.86MB |
| Vocal SDR | 8.71dB | 6.53dB | 6.29dB |
| BGM SDR | 14.70dB | 12.84dB | 12.52dB |
SDR (Signal-to-Distortion Ratio) shows how clean a split is — higher means less bleed. NPU 100% / 25% — time when using all / a quarter of the device’s AI chip (NPU).
Key Features
Just the voice, cleanly.
- 01
Vocal · BGM split
Splits a mix into vocals and background music (BGM).
- 02
3 models
Pick Lite for quality, Compact for balance or Nano for speed.
- 03
NPU-accelerated
Runs on the device’s AI chip (NPU). Nano needs just 8.0ms.
- 04
Light library
From 10.86MB — fits mobile and embedded devices.
Specifications
RASES VC at a glance.
- Split
- 2 — vocals, BGM
- Models
- Lite · Compact · Nano
- Time (NPU 100%)
- 8.0 – 26.1ms
- Time (NPU 25%)
- 32.2 – 104.3ms
- Library
- 10.86 – 22.21MB
- Quality (SDR)
- Vocal 6.29 – 8.71dB · BGM 12.52 – 14.70dB
Custom development available
Let’s build
your sound.
We co-develop immersive audio tailored to your product and platform.
Choose a product to get in touch.