Sign inSign up
split-my-music logo

paladini/split-my-music

By paladini

•Updated 1 day ago

Self-hosted AI vocal, music & instrument separation with a friendly web UI (CPU & NVIDIA GPU).

Image
0

170

paladini/split-my-music repository overview

⁠Split My Music

Self-hosted web application that separates vocals, music, drums, bass, guitar, and piano from your audio files using open-source AI models via python-audio-separator⁠. Everything runs locally on your own server — no external audio uploads.

Source code, documentation, and issue tracker: https://github.com/paladini/split-my-music⁠

⁠Quick Start (CPU — Tested & Ready)

docker run -d \
  --name split-my-music \
  -p 7860:7860 \
  -v split-models:/data/models \
  -v split-jobs:/data/jobs \
  --restart unless-stopped \
  paladini/split-my-music:latest

Open http://localhost:7860. Models download once on first use (~50–80 MB for default models) and persist in /data/models.

⁠NVIDIA GPU

Requires an NVIDIA GPU, recent NVIDIA host driver, and the NVIDIA Container Toolkit:

docker run -d \
  --name split-my-music \
  --gpus all \
  -p 7860:7860 \
  -v split-models:/data/models \
  -v split-jobs:/data/jobs \
  --restart unless-stopped \
  paladini/split-my-music:gpu

Note: The CPU image was built and smoke-tested locally; the GPU image is built via CI/CD from Dockerfile.gpu and has not yet been executed on physical NVIDIA hardware by the author. Apple Silicon, AMD, and Intel GPUs are not accelerated inside these Linux containers.

⁠Available Tags

TagVariantArchitecture
latest, cpu, 1.0.0, 1.0CPU (Dockerfile)linux/amd64
gpu, 1.0.0-gpu, 1.0-gpuNVIDIA CUDA 12.6 (Dockerfile.gpu)linux/amd64

⁠Configuration

  • SPLIT_MAX_UPLOAD_MB (200)
  • SPLIT_MAX_DURATION_MINUTES (20)
  • SPLIT_MAX_CONCURRENT_JOBS (1)
  • SPLIT_MAX_QUEUE_SIZE (5)
  • SPLIT_MIN_FREE_DISK_MB (500)
  • SPLIT_RETENTION_HOURS (24)
  • SPLIT_MAX_STORED_JOBS (20)
  • SPLIT_LOG_LEVEL (INFO)

Health check endpoint: GET /health

⁠Security

Split My Music ships without login/authentication and is intended for local machines, home LANs, and homelabs. Do not expose it directly to the public internet without placing it behind an authenticated HTTPS reverse proxy.

⁠License

MIT License for Split My Music application code. Model weights are downloaded at runtime and retain their upstream licenses; see THIRD_PARTY.md in the repository.

Tag summary

Content type

Image

Digest

sha256:103b323a6…

Size

684.6 MB

Last updated

1 day ago

docker pull paladini/split-my-music