Return HiFi-GAN vocoder

app.py

@@ -15,7 +15,7 @@ asr_pipe = pipeline("automatic-speech-recognition", model="openai/whisper-base",
 tts_model_name = "sanchit-gandhi/speecht5_tts_vox_nl"
 processor = SpeechT5Processor.from_pretrained(tts_model_name)
 model = SpeechT5ForTextToSpeech.from_pretrained(tts_model_name).to(device)
-vocoder = SpeechT5HifiGan.from_pretrained(tts_model_name).to(device)
+vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan").to(device)
 
 embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
 speaker_embeddings = torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze(0)