roformer

Production-ready UVR5 CLI & Docker image. Run SOTA separation models (Roformer, SCNet, MDX, Demucs, VR Architecture) on headless GPU servers without dependency hell.

MLX-native audio stem separation for Apple Silicon (Demucs, MDX, MDXC, VR, Roformer).

This repository contains all public UVR models, with possibly a few private ones included.

Clean audiobooks from background music and noise using MDX-Net/Roformer AI models. Handles large files with automatic chunking.

群友让我优化一下UVR5的4个常用的模型的推理速度，以及运行时环境的大小

Split any song into vocals and instrumental, entirely on your own computer. Free, open source, works offline.

VocalDrop is a free AI vocal isolator and remover for macOS, Windows and Linux that runs entirely on your own hardware. No upload, no watermark, no subscription.

Remove vocals from any song: karaoke instrumental and isolated vocals. Demucs + FastAPI + Svelte.

Remove the electric guitar from any song to make a backing track (MelBand-Roformer)

Split a song into 13 true-level stems (vocals, drum-kit pieces, bass, keys, guitar) that sum back to the mix.

AI-powered desktop DAW - stem separation, chord detection, karaoke

Turn any track into a DJ stem file (.stem.mp4) for Mixxx with state-of-the-art open separation models

Local Stem AI — local GPU audio separation studio

Local AI stem separation for music producers. Private processing, batch queue, WAV preview and DAW drag-out. Linux-first.

Separazione stem (Demucs/Roformer) + mixer di pratica, testi e accordatore, per Windows. | Stem separation + practice mixer, lyrics & tuner for Windows.

本地人声/伴奏分离工作站 · 全离线、GPU 推理、去回声 RVC 前置 | Local vocal/instrumental separator, fully offline with built-in models

Senior Year Capstone Project

Pull songs apart into vocals, drums and more on your own device, with the best available separation models

Split songs into vocals and instrumental with Mel-Band RoFormer. One command, GPU-accelerated, handles single files or whole folders.

A from-scratch study of Rotary Position Embedding in Transformers, including implementation, ablation experiments, language modeling evaluation, and comparison with the original RoFormer paper.