Skip to content

v0.4.0

Choose a tag to compare

View all tags
@danielr-ceva danielr-ceva released this 20 Mar 20:08
· 22 commits to main since this release
v0.4.0
2da0a2d

v0.4.0 Release Notes

Highlights

This release focuses on ONNX export quality, real-time audio processing, and broader audio format support.

What's New

ONNX Export & Inference Improvements
Grouped linear layers now export as clean single MatMul nodes, with automatic graph simplification via onnxsim and full ORT optimization enabled across all inference sessions - resulting in faster and leaner ONNX models.

Real-Time Audio Processing
A new StreamEnhancer class enables low-latency, chunk-by-chunk audio enhancement for live/streaming use cases.

Multi-Format Audio Input
The enhancer now accepts .flac, .ogg, .aiff natively, plus .mp3, .m4a, .aac, .wma, and .opus via an optional pydub+ffmpeg dependency. Output remains PCM16 .wav.

Concurrent Directory Enhancement
enhance-dir now supports a --workers flag for parallel processing of audio files.

Other Changes

  • Cleaner CLI models table formatting
  • Expanded test coverage across audio, models, API, CLI, and streaming
  • Removed .npz state export in favor of ONNX metadata
  • README updates
Assets 2
Loading