Update app.py

app.py

@@ -12,7 +12,7 @@ device = "cuda:0" if torch.cuda.is_available() else "cpu"
 asr_pipe = pipeline("automatic-speech-recognition", model="openai/whisper-base", device=device)
 
 # load text-to-speech checkpoint and speaker embeddings
-model_name = "wuula/speecht5_tts_common_voice_zh"
+model_name = "sanchit-gandhi/speecht5_tts_vox_nl"
 processor = SpeechT5Processor.from_pretrained(model_name)
 
 model = SpeechT5ForTextToSpeech.from_pretrained(model_name).to(device)
@@ -23,7 +23,7 @@ speaker_embeddings = torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze
 
 
 def translate(audio):
-    outputs = asr_pipe(audio, max_new_tokens=256, generate_kwargs={"task": "transcribe", "language": "zh"})
+    outputs = asr_pipe(audio, max_new_tokens=256, generate_kwargs={"task": "transcribe", "language": "nl"})
     # outputs = asr_pipe(audio, max_new_tokens=256, generate_kwargs={"task": "transcribe"})
     return outputs["text"]