KinWaiCheuk/nnAudio

Health v2 · maintenance only

Adoption not part of the score

1127 stars · 97 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

nnAudio is a PyTorch-based audio processing toolbox that computes spectrograms (STFT, Mel, CQT, MFCC) using 1D convolutional neural networks. This enables on-the-fly spectrogram generation during training and makes the Fourier/CQT kernels trainable and differentiable.

Use cases

• compute spectrograms on gpu during neural network training
• trainable cqt kernels in pytorch
• differentiable mel spectrogram for deep learning
• audio preprocessing inside a pytorch model
• stft and inverse stft with griffin-lim in pytorch
• cross-platform alternative to torchaudio on windows

When to choose

• you need trainable or differentiable spectrogram front-ends in PyTorch
• you want fast GPU-accelerated audio feature extraction during training
• torchaudio installation is problematic on your platform

When to avoid

• you only need offline numpy-based analysis (use librosa)
• you need a actively developed project - the successor nnAudio2 is maintained instead
• you work outside the PyTorch ecosystem

Facets

library · maturity maintenance

audio-processing machine-learning deep-learning machine-learning deep-learning python cross-platform pytorch spectrogram stft mel-spectrogram cqt trainable-kernels gpu-audio audio gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="KinWaiCheuk/nnAudio")