Split Model

Youka offers different AI models for separating vocals from instrumentals. Choose the right one for your needs.

Available Models #

MDX-23C (Default) #

High-quality vocal separation:

• Instrumental: Music without vocals
• Lead Vocals: Main singer

Best for: Most karaoke applications

AudioShakeAI #

Professional-grade separation:

• Instrumental: Music without vocals
• Lead Vocals: Main singer

Best for: High-quality results

MusicAI (Instrumental Only) #

Only extracts the instrumental:

• Instrumental: Music without any vocals

Best for: When you only need the backing track

MusicAI (Instrumental / Backing / Lead) #

The only model that produces 3 separate stems:

• Instrumental: Music only
• Backing Vocals: Background singers, choirs
• Lead Vocals: Main singer

Best for: Full control over backing vocals mixing

MusicAI (With Backing Vocals) #

Keeps backing vocals with the instrumental:

• Instrumental + Backing Vocals: Music with background singers
• Lead Vocals: Main singer only

Best for: Karaoke where you want choir/harmony support

MusicAI (Without Backing Vocals) #

Removes all vocals including backing:

• Instrumental: Music without any vocals
• Lead Vocals: Main singer

Best for: Clean instrumental without any voices

MDX-Net (With Backing Vocals) #

Alternative model that keeps backing vocals with instrumental:

• Instrumental + Backing Vocals: Music with background singers
• Lead Vocals: Main singer only

Best for: Alternative to MusicAI backing vocals model

RoFormer Models #

Various RoFormer-based models for vocal separation:

• RoFormer: Standard RoFormer model
• RoFormer (becruily): Tuned variant
• RoFormer (instv7_gabox): Another tuned variant

Best for: Trying different separation quality

Demucs #

Open-source separation model:

• Instrumental: Music without vocals
• Lead Vocals: Main singer

Best for: Alternative when other models don't work well

Choosing a Model When Creating Karaoke #

1. Click Create Karaoke
2. Expand the advanced options
3. Select your preferred split model
4. Complete the creation process

Creating New Stems for Existing Projects #

To try a different model on an existing project:

1. Open the project
2. Go to the Audio menu
3. Under Create, click on the model you want to try
4. Wait for processing to complete
5. The new stems will appear in the Audio menu