Speech Enhancement Example

This example shows how to use the Dart offline speech denoiser API with DPDFNet models.

Use 16 kHz DPDFNet models such as dpdfnet_baseline.onnx, dpdfnet2.onnx, dpdfnet4.onnx, or dpdfnet8.onnx for downstream ASR or speech recognition. Use dpdfnet2_48khz_hr.onnx for 48 kHz enhancement output.

DPDFNet models are available from either:

• https://github.com/k2-fsa/sherpa-onnx/releases/tag/speech-enhancement-models
• https://huggingface.co/Ceva-IP/DPDFNet

Then run:

dart pub get
dart run ./bin/speech_enhancement_dpdfnet.dart --model ./dpdfnet_baseline.onnx --input-wav ./inp_16k.wav --output-wav ./enhanced-16k.wav