STEM STUDIO
AI vocal remover & stem splitter, 100% in your browser
Drop in any track and split it into clean vocal and instrumental stems—right inside your browser. Stem Studio runs a real AI source-separation model (UVR-MDX-NET) on-device with ONNX Runtime Web, so your audio is never uploaded and there's no server round-trip. Mix each stem live with independent volume and mute controls, then download the results as WAV. Today it separates two stems—vocals and instrumental—with more stem types on the roadmap.
FEATURES
- AI Vocal Remover / Instrumental Splitter
- Runs 100% On-Device — No Upload, No Server
- Drag & Drop Upload
- Per-Stem Volume & Mute
- Waveform Display
- WAV Stem Download
- CPU (WASM) or GPU (WebGPU) Engine
- Optional Denoise Pass
- Model Caching — Downloads Once, Instant After
*Separation runs entirely in your browser — your audio is never uploaded to a server
*First run downloads the ~64MB separation model (cached afterward for instant reuse)
EXTERNAL LIBRARIES
22.1.4 Frequently Asked Questions
Are my audio files uploaded to a server?
No. Stem Studio runs the UVR-MDX-NET separation model on-device with ONNX Runtime Web — your audio is never uploaded and there's no server round-trip.
How many stems can Stem Studio separate?
Today it separates two stems — vocals and instrumental — with more stem types on the roadmap.
Does Stem Studio use my CPU or GPU?
Either. You can run separation on CPU via WASM or on GPU via WebGPU. The ~64MB model downloads once and is cached for instant reuse afterward.
What do I get when I buy repo access?
Read-only (pull) access to the Stem Studio private GitHub repository, for one year by default or for life with the lifetime option (twice the price). A GitHub account is required to receive the invite.