search

hayeong0 / Diff-HierVC

Official Pytorch Implementation of "Diff-HierVC: Diffusion-based Hierarchical Voice Conversion with Robust Pitch Generation and Masked Prior for Zero-shot Speaker Adaptation"
https://diff-hiervc.github.io/
189 stars 18 forks source link

bigvgan #4

Closed bigdan12 closed 8 months ago

bigdan12 commented 8 months ago

The config may not be correct, please update it.

hayeong0 commented 8 months ago

I've updated it now. Please check it.

weirenlan commented 8 months ago

@hayeong0 Thanks for your research I try to inference with bigvan by this setting 

bash infer.sh

python3 inference.py \
    --src_path './sample/src_p241_004.wav' \
    --trg_path './sample/tar_p239_022.wav' \
    --ckpt_model './ckpt/model_diffhier.pth' \
    --ckpt_voc './vocoder/voc_bigvgan.pth' \
    --output_dir './converted' \
    --diffpitch_ts 30 \
    --diffvoice_ts 6

and adjust the config with

        config = os.path.join(os.path.split(a.ckpt_model)[0], 'config_bigvgan.json')

But the converted result seems like strange

image
hayeong0 commented 8 months ago

@weirenlan

We didn't encounter any issues when performing inference with our existing code. The results you posted seem to indicate that Bigvgan has not been loaded. I have made some simplifications to the code to make it easier. Please try again with the updated code and make sure to check the location where you downloaded the checkpoint file :)

weirenlan commented 8 months ago

@hayeong0 Thanks for your updating. After the updating, the export is correct and awesome~