README.md

metadata

title: Demucs Stem Separator
emoji: 🎵
colorFrom: purple
colorTo: pink
sdk: gradio
sdk_version: 6.3.0
python_version: '3.10'
app_file: app.py
pinned: false
license: mit
tags:
  - audio-separation
  - stem-separation
  - demucs
  - mcp-server
short_description: Separate audio into vocals, drums, bass, other stems

🎵 Demucs Stem Separator

Separate audio → Vocals | Drums | Bass | Other

Stem	Output
🎤 Vocals	Singing, voice
🥁 Drums	Percussion
🎸 Bass	Low frequencies
🎹 Other	Guitars, synths

Formats: MP3, WAV, FLAC, OGG, M4A

---

API

Python Client

from gradio_client import Client, handle_file

client = Client("Luminia/demucs")
vocals, drums, bass, other = client.predict(
    audio_file=handle_file("song.wav"),
    model_name="htdemucs",  # or "htdemucs_ft"
    api_name="/separate"
)

REST API (curl)

# Submit
curl -X POST "https://luminia-demucs.hf.space/gradio_api/call/separate" \
  -H "Content-Type: application/json" \
  -d '{"data": ["https://example.com/song.mp3", "htdemucs"]}'

# Get result (use event_id from response)
curl "https://luminia-demucs.hf.space/gradio_api/call/separate/{event_id}"

MCP

{
  "mcpServers": {
    "demucs": {
      "url": "https://luminia-demucs.hf.space/gradio_api/mcp/"
    }
  }
}

---

Demucs by Facebook Research