---
title: Sever
emoji: "✂️"
colorFrom: gray
colorTo: purple
sdk: docker
app_port: 7860
pinned: false
---

# Sever

HTDemucs-6s stem separation for BASE Station. Six stems: drums, bass, other,
vocals, guitar, piano.

Asynchronous by necessity — separation takes minutes on CPU.

- `POST /separate` → `{ audio_url, stems? }` → `{ job_id }`
- `GET /status/{job_id}` → `{ status, stems: { name: "/outputs/..." } }`
- `GET /health`

Jobs run one at a time on purpose: concurrent separations each hold a full
model plus a full audio tensor in RAM and reliably OOM.