Samples and pretrained models

[jmlemercier]

Hi, thank you for the upload and the paper. Do you have by any chance some pretrained models lying around for us to experiment with, and also some audio examples or your approach, as compared to CDiffuse and other baselines?

Many thanks!

[zhibinQiu]

@jmlemercier Thank you for your interest in my work. I will upload the pre-trained model as well as the enhanced samples soon. Also due to the rush, there may be some errors in the current code, I plan to reorganize the code and upload it to the current repository, please stay tuned for updates. Thank you again!

[jmlemercier]

All good, thanks for maintaining!

[jmlemercier]

Thanks for uploading the code and the pretrained models. I managed to make the code work, but I think that if you want people to use your code to compare and cite your study, you should consider cleaning and testing your code repo:

• many filenames are wrong (model_inital_predictor --> model_initial_predictor) , 'model_alpha_no_bar_stft_refine.py' does not exist and should be replaced by 'model.py' to import DiffuSE, ...)
• the inference script does not work out of the box, I had to rewrite and simplify it (why would you use the 'spectrogram' argument to infer the shapes? You just have to pad by the next integer noisy_signal.shape[0]//model.params.hop_samples + 1). Also your model.params.sample_rate returns 10000 which should be 16000. That makes the audio output incorrect.