Error when using fairseq models about sub-to-audio

[bnsantoso]

if you want use fairseq vietnam model : sub = SubToAudio(language='vie') .

sub = SubToAudio(languages='vie') 
subtitle = sub.subtitle("01.vi.srt")
sub.convert_to_audio(data=subtitle, tempo_mode="all", tempo_speed=1.3, output_path="01.vi.wav", save_temp=True,)

there is no need to pass languages argument on self.convert_to_audio() func. fairseq model only have 1 speaker and 1 languages. and i notice that you give wrong argument lang="vie", instead languages="vie".

arg languages in SubToAudio() and self.convert_to_audio() have different purpose. SubToAudio(languages=<iso-code>) is for initialze and downloading model. self.convert_to_audio(languages="fr-fr") is for specify the languages within model that have multiple languages.

if you have your own model and config sub = SubToAudio(model_path="G_100000.pth" , config_path="config.json"), make sure it is correct model and corect config and correct path. isolate class instance to test or you can directly test your model from TTS api to check wether is perfeclty fine and can load.

maybe you have similiar issue like this, the model and config.json is not suported, wrong or does not match with model :
AttributeError: 'dict' object has no attribute 'replace'

if the model have multiple speaker and languages you can check them print(sub.apitts.languages) and print(sub.apitts.speakers).

from subtoaudio import SubToAudio
sub = SubToAudio(model_path="G_100000.pth" , config_path="config.json")
print(sub.apitts.languages) #print list of language on model
print(sub.apitts.speakers) #print list of speakers