Use Dutch language model for TSS

app.py

@@ -13,11 +13,13 @@ asr_pipe = pipeline("automatic-speech-recognition", model="openai/whisper-base",
 
 # load text-to-speech checkpoint and speaker embeddings 
 #processor = SpeechT5Processor.from_pretrained("microsoft/speecht5_tts")
-processor = SpeechT5Processor.from_pretrained("ihanif/speecht5_finetuned_voxpopuli_lt")
+#"ihanif/speecht5_finetuned_voxpopuli_lt"
+model_id = "sanchit-gandhi/speecht5_tts_vox_nl"
+processor = SpeechT5Processor.from_pretrained(model_id)
 
 #model = SpeechT5ForTextToSpeech.from_pretrained("microsoft/speecht5_tts").to(device)
 #vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan").to(device)
-model = SpeechT5ForTextToSpeech.from_pretrained("ihanif/speecht5_finetuned_voxpopuli_lt").to(device)
+model = SpeechT5ForTextToSpeech.from_pretrained(model_id).to(device)
 vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan").to(device)
 
 embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
@@ -25,7 +27,7 @@ speaker_embeddings = torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze
 
 
 def translate(audio):
-    outputs = asr_pipe(audio, max_new_tokens=256, generate_kwargs={"task": "translate", "language": "lt"})
+    outputs = asr_pipe(audio, max_new_tokens=256, generate_kwargs={"task": "translate", "language": "nl"})
     return outputs["text"]