Documentation

Everything you need to get started with Splitvoice.

Getting Started with Splitvoice

Follow these simple steps to start separating stems.

1. Install Splitvoice

Download the latest version for Windows from the Download page. Run the installer and follow the setup wizard.

2. Run First-Time Setup

On first launch, Splitvoice will walk you through installing Python and downloading the AI models. This one-time setup downloads approximately 2 GB of model data.

3. Select an Audio File

Click 'Browse File' or drag and drop an audio file onto the interface. Splitvoice supports MP3, WAV, FLAC, M4A, and OGG formats.

4. Choose Your Model

Free includes Demucs (4 stems). Pro adds BS-RoFormer (6 stems: + guitar, piano).

5. Start Splitting

Click 'Split Audio Stems' and wait for the process to complete. Processing time varies based on file length and hardware.

6. Preview & Export

Once complete, use the Stem Mixer to preview individual stems or create custom mixes. Exported files are saved to your chosen output directory.

Available AI Models

Splitvoice comes with two powerful AI models for stem separation (BS-RoFormer available in Pro).

Demucs (htdemucs_ft) (Free / Pro)

Meta's Hybrid Transformer Demucs model, fine-tuned for optimal quality. Produces 4 stems: vocals, drums, bass, and other. This is the default model and works well for most music.

BS-RoFormer-SW (Pro)

An advanced spectrogram-based model that produces 6 stems: vocals, drums, bass, guitar, piano, and other. Offers finer instrument separation at the cost of slightly longer processing time.

Pro Tips

For best results with stem separation:

  • Use high-quality source files (WAV or FLAC at 44.1kHz+)
  • Enable GPU acceleration if you have a compatible NVIDIA GPU
  • For karaoke, mix vocals stem at -∞ and export
  • Use BS-RoFormer for tracks with prominent piano or guitar
  • Keep your output directory organized by using different folders for different projects