Releases: Rikorose/DeepFilterNet
Releases · Rikorose/DeepFilterNet
v0.5.6
v0.5.4
This release adds a few more python wheels to PyPI
- Add python 3.11 support
- Add MacOS aarch64
- Add Linux aarch64
v0.5.3
- Fix a bug where the ladspa plugin results in a reverberant output when running with attenuation limit
v0.5.2
Minor release with improved logging.
v0.5.1
Bugfix release
- Better ladspa defaults for consistent noise suppression
- Revert DNN inference framework to previous version due to audible artifacts
v0.5.0
- Many bugfixes
- Improved DeepFilterNet3 model
- Multi-Frame Filtering (MVDR/Wiener Filter)
v0.4.0
- New DeepFilterNet3 model with a slightly improved noise suppression performance.
- Fix a bug, where the native implementation produced a slightly worse noise suppression compared to the pytorch implementation.
- Add attenuation limiting for the LADSPA plugin.
v0.3.1
- Fix a regression that caused a larger memory consumption
- Related to above: Update onnx models for deep-filter Rust implementation and improve inference speed
0.3.0
Rust-only version of DeepFilterNet implemented as a real-time loop.
This makes processing of large files relatively easy. Usage:
$ deep-filter -h
deep_filter 0.3.0
Hendrik Schröter
Noise supression using deep filtering
USAGE:
deep-filter [OPTIONS] [FILES]...
ARGS:
<FILES>...
OPTIONS:
-D, --compensate-delay
Compensate delay of STFT and model lookahead
-h, --help
Print help information
-m, --model <MODEL>
Path to model tar.gz
-o, --out-dir <OUT_DIR>
[default: out]
--pf
Enable postfilter
--reduce-mask <REDUCE_MASK>
If used with multiple channels, reduce the mask with max (1) or mean (2) [default: 1]
-v, --verbose
Logging verbosity
-V, --version
Print version information
LADSPA plugin for real-time noise reduction
E.g. for usage in a Pipewire filter chain. See readme.