Update app.py

app.py

@@ -12,9 +12,9 @@ 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
-processor = SpeechT5Processor.from_pretrained("catactivationsound/speecht5_finetuned_voxpopuli_hu")
+processor = SpeechT5Processor.from_pretrained("sanchit-gandhi/speecht5_tts_vox_nl")
 
-model = SpeechT5ForTextToSpeech.from_pretrained("catactivationsound/speecht5_finetuned_voxpopuli_hu").to(device)
+model = SpeechT5ForTextToSpeech.from_pretrained("sanchit-gandhi/speecht5_tts_vox_nl").to(device)
 vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan").to(device)
 
 embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
@@ -22,7 +22,7 @@ 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": "hu"})
+    outputs = asr_pipe(audio, max_new_tokens=256, generate_kwargs={"task": "transcribe", "language": "nl"})
     return outputs["text"]
 
 
@@ -41,9 +41,9 @@ def speech_to_speech_translation(audio):
 
 title = "Cascaded STST"
 description = """
-Demo for cascaded speech-to-speech translation (STST), mapping from source speech in any language to target speech in Hungarian.
+Demo for cascaded speech-to-speech translation (STST), mapping from source speech in any language to target speech in Dutch.
 Demo uses OpenAI's [Whisper Base](https://huggingface.co/openai/whisper-base) model for speech translation,
-and a [finetuned version](https://huggingface.co/catactivationsound/speecht5_finetuned_voxpopuli_hu)
+and a [finetuned version](https://huggingface.co/sanchit-gandhi/speecht5_tts_vox_nl)
 of Microsoft's [SpeechT5 TTS](https://huggingface.co/microsoft/speecht5_tts) for text-to-speech:
 
 ![Cascaded STST](https://huggingface.co/datasets/huggingface-course/audio-course-images/resolve/main/s2st_cascaded.png "Diagram of cascaded speech to speech translation")