About Opus

Opus is a free, browser-based AI vocal remover and karaoke maker. Upload any MP3 and get back a clean instrumental in under a minute — no account, no watermark, no paywall.

Why it exists

Most vocal removers online are either low-quality (old center-channel tricks), locked behind subscriptions, or require signup and give you watermarked 30-second clips. Opus was built because none of that should be necessary for something a neural network can do in seconds.

The goal is simple: give musicians, karaoke hosts, teachers, and hobbyists a tool that respects their time, their files, and their wallet.

How it works

Opususes Demucs — a state-of-the-art source separation model from Meta's FAIR research lab — to split audio into stems: vocals, drums, bass, and other instruments. The instrumental you download is the sum of everything except the vocals.

Processing runs on server GPUs so your browser and device stay responsive. Files are deleted within an hour of processing, and we don't train any model on your audio.

What you can do with it

• Make karaoke tracks from any song
• Extract instrumentals for remixes, mashups, and sampling
• Isolate backing music for practice or transcription
• Detect tempo, key, and time signature (Western BPM and Indian tala)

Who built it

Opus is built and maintained by Sathvik SVS, an independent developer. It's a solo project — so bug reports, ideas, and feedback land directly in the inbox of the person who can act on them.

Contact

Bug, feature request, or just want to say hi? Email sathvik@hashiraworks.com or visit the contact page.