An Open-Source Project to Unify Audio Processing and Generation

Official implemtation of UniverSR (ICASSP 2026)

🩺🎧 Fix all your podcast, video or live stream audio! 🎧🩺

Mp3 to wav super resolution model for audio restoration & enhancement. U-Net + Discrete Wavelet Transform (DWT) Architecture

bin2bin, a Time-Frequency Generative Adversarial based method for Audio Packet Loss Concealment

Audio restoration platform implementing 1D Operational GANs (Kiranyaz et al. 2022) for blind denoising of degraded recordings. Production API uses UVR for source separation; experimental OpGAN pipeline for analog noise removal. Full-stack: FastAPI, React, Docker.

iZotope RX is the industry-standard audio repair suite with spectral editing, dialogue de-noise, de-reverb, de-clip, and breath control. Used in music, film, and podcast production.

A research based project which uses steganography and ML/deep learning algorithm to reconstruct the lost audio signals from a corrupted file.

One-click setup for iZotope RX Audio Repair Setup -- the industry-standard AI audio repair suite used in film, TV, and podcast production to remove noise, hum, reverb, and background voices from any recording

A high-performance C++ port of HRAudioWizard - an audio enhancement tool that uses high-frequency compensation (HFC) to improve the quality of low-fidelity recordings by synthesizing plausible high-frequency content.

Short-Time Spectral Attenuator, plus attempt to reverse the process in order to have a noise-based music effect

A studio-grade, fully automated VHS audio restoration suite. Uses a hybrid AI engine to separate and clean vocals while preserving background music and singing voices. Includes RTX 5090 fallback support.

A high-performance local AI pipeline for restoring VHS audio, transcribing with Whisper, and translating subtitles using NLLB-200.

Code and demo site for complex-bin2bin speech signals inpainting

Local AI audio restoration. Phone recording to podcast quality. Zero cloud.

McGill University - Digital Audio Signal Processing

It is a professional-grade digital audio workstation designed for precision waveform editing, complex multitrack mixing, and industry-leading audio restoration for film, broadcast, and podcasting

Restore, enhance and level the audio in any recording — in pure Rust

AI-powered audio repair, restoration & stem-remix rack plugin — VST3 / AU / LV2 / Standalone (JUCE + ONNX Runtime)

A full-stack web application for removing impulsive noise (clicks, pops, and crackle) from audio recordings using adaptive Autoregressive (AR) modeling and Exponentially Weighted Least Squares (EW-LS) parameter estimation.