Model Information

Display Name: Ryan Demucs

API Model ID: ryan/demucs

Category: Audio To Audio

Description: Demucs v4 audio source separation. Splits any song into individual stems: vocals, drums, bass, and other instruments. Optionally separates guitar and piano with the 6-stem model. ## Use Cases - **Isolate vocals** for voice conversion with pseudoram/rvc-v2 - **Remove vocals** to create instrumentals/karaoke - **Extract drums/bass** for remixing - **Full stem separation** for music production ## Models - **auto**: Automatically picks the best model - **htdemucs**: Base hybrid transformer (fast) - **htdemucs_ft**: Fine-tuned, best quality (~4x slower) - **htdemucs_6s**: 6 stems (adds guitar + piano)

How to Use This Model

To use Ryan Demucs via the HInow.ai API, use the model ID: ryan/demucs

API Request Example

POST https://api.hinow.ai/v1/chat/completions
Authorization: Bearer YOUR_API_KEY
Content-Type: application/json

{
  "model": "ryan/demucs",
  "messages": [
    {"role": "user", "content": "Your input here"}
  ]
}
              

Pricing

• total: $0.05

Available Parameters

• audio: Audio file URL to separate (MP3, WAV, FLAC)
• model: Separation model. auto=best for task, htdemucs_ft=highest quality. Default: auto (Options: auto, htdemucs, htdemucs_ft, htdemucs_6s)
• isolate_stem: Isolate a single stem (returns stem + no_stem). Default: none (all stems) (Options: none, vocals, drums, bass, other, guitar, piano)
• format: Output audio format. Default: wav (Options: wav, mp3, flac)
• shifts: Random shifts for quality stabilization 1-20. More=better but slower. Default: 1
• split_overlap: Segment overlap 0-1. Higher=better at boundaries. Default: 0.25
• clip_mode: Clipping prevention method. Default: rescale (Options: rescale, clamp, tanh, none)

Quick Reference

To use this model, set: "model": "ryan/demucs"

Featured: No

Documentation: https://hinow.ai/models/ryan/demucs

API Endpoint: https://api.hinow.ai/v1