|
|
|
|
@ -312,6 +312,7 @@ class TextToSpeech:
|
|
|
|
|
self.cvvp = None # CVVP model is only loaded if used.
|
|
|
|
|
|
|
|
|
|
if use_bigvgan:
|
|
|
|
|
# credit https://github.com/deviandiceto / https://git.ecker.tech/mrq/ai-voice-cloning/issues/52
|
|
|
|
|
self.vocoder = BigVGAN().cpu()
|
|
|
|
|
state_dict_bigvgan = load_checkpoint(get_model_path('bigvgan_base_24khz_100band.pth', models_dir), self.device)
|
|
|
|
|
self.vocoder.load_state_dict(state_dict_bigvgan['generator'])
|
|
|
|
|
|
