Cathar
github.com
- Free plan
- macOS
- Linux
What Cathar is
Cathar is a free audio restoration toolkit for Linux and macOS. It cleans and enhances recordings, then provides loudness and peak normalization to prepare output levels. It accepts WAV, MP3, FLAC, OGG, and M4A files, along with audio tracks inside MP4 and MKV video, and writes WAV, FLAC, or AIFF according to the output filename extension. Restoration tools cover noise, hum, wind, clicks, clipping, reverb, plosives, rustle, and spectral repair. Enhancement options include de-essing, breath removal, voice isolation, and bandwidth extension. Written in Rust, Cathar uses processing stages that can be inspected and tuned rather than opaque neural models. Effects can run in-process through the Rust crate or as a command-line stage in a media pipeline. An optional feature adds a deterministic GRU denoiser using safetensors weights. Optional terminal features provide an audio player and interactive spectrogram viewer; Linux playback requires ALSA headers. The project is licensed under Apache License 2.0 or MIT and is in its 0.8.x series before 1.0.
Who it suits
Cathar suits audio editors who want a free toolkit for cleaning and leveling recordings on Linux or macOS. Its Rust crate and command-line interface also suit developers building audio-processing pipelines.
Good to know, and what to watch
- Accepts audio tracks inside MP4 and MKV files.
- Includes spectral repair and de-clip tools.
- Effects can run in-process or in a pipeline.
- Optional feature provides a GRU-based denoiser.
- Licensed under Apache 2.0 or MIT.
- Supports Linux and macOS only.
- Linux terminal playback requires ALSA headers.
- The project is still before version 1.0.
The verdict
Cathar brings a wide set of restoration and enhancement functions into a free, Rust-based toolkit. Its pre-1.0 status and Linux playback requirement are worth considering for production workflows.
What the maker publishes
- Current release
- The GitHub releases page lists cathar v0.8.0 as the latest release. source
- Documentation
- The project includes an illustrated Cleaning Up Sound book explaining its concepts and comparing it with tools such as iZotope RX, Audition, Audacity, SoX and FFmpeg. source
- DSP approach
- Its processing stages are inspectable and tunable DSP rather than opaque neural models. source
- Enhancement
- Enhancement tools include de-essing, breath removal, voice isolation and bandwidth extension. source
- enhancement
- Its enhancement functions include de-essing, breath removal, voice isolation, and bandwidth extension. source
- formats
- It reads WAV, MP3, FLAC, OGG, M4A, MP4, and MKV inputs and writes WAV, FLAC, or AIFF. source
- Implementation
- The toolkit is written in pure Rust and uses no ffmpeg, C or C++ code, or system libraries. source
- implementation
- Cathar is implemented in pure Rust with no ffmpeg, C/C++, or system-library dependencies. source
- Input formats
- It accepts WAV, MP3, FLAC, OGG and M4A files, plus audio tracks inside MP4 and MKV video. source
- Installation
- The command-line binary can be installed with cargo install cathar-cli from crates.io. source
- integration
- The toolkit can be called in-process through the cathar Rust crate or invoked as a command-line stage in a media pipeline. source
- latest_release
- The GitHub releases page lists cathar v0.8.0 as the latest release. source
- Leveling
- Cathar provides LUFS loudness normalization and peak normalization for delivery. source
- leveling
- Cathar supports LUFS and peak normalization for delivery. source
- License
- Cathar is licensed under either the Apache License 2.0 or the MIT License. source
- license
- The project is licensed under either Apache License 2.0 or MIT. source
- Machine-learning denoise
- An optional ml feature adds a GRU-based denoiser whose open safetensors weights and deterministic inference are implemented in Rust. source
- maker
- The repository owner is Vikram Bhaskaran, whose profile describes him as a technologist and lists Sydney. source
- ml_denoise
- An optional ml feature provides a deterministic GRU-based denoiser that loads safetensors checkpoints. source
- Output formats
- It writes WAV, FLAC or AIFF output selected by the output filename extension. source
- Pipeline integration
- Effects are plain functions over audio sample slices and can run in-process through the cathar crate or as a cathar command between pipeline steps. source
- Purpose
- Cathar is an audio restoration toolkit for cleaning, enhancing and leveling recordings. source
- purpose
- Cathar is an audio restoration toolkit for noisy recordings and video audio tracks. source
- Restoration
- Restoration tools include denoise, de-hum, de-wind, de-click, de-clip, de-reverb, spectral repair, de-plosive and de-rustle. source
Checked against the maker's pages on 2 Oct 2026.