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("adavirro/speecht5_finetuned_voxpopuli_it")
+processor = SpeechT5Processor.from_pretrained("hlumin/speecht5_finetuned_voxpopuli_nl")
 
-model = SpeechT5ForTextToSpeech.from_pretrained("adavirro/speecht5_finetuned_voxpopuli_it").to(device)
+model = SpeechT5ForTextToSpeech.from_pretrained("hlumin/speecht5_finetuned_voxpopuli_nl").to(device)
 vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan").to(device)
 
 embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
@@ -39,11 +39,11 @@ def speech_to_speech_translation(audio):
     return 16000, synthesised_speech
 
 
-title = "STST in LoFi Italian 😎"
+title = "STST - Lithuanian"
 description = """
-Demo for cascaded speech-to-speech translation (STST), mapping from source speech in any language to super lo-fi Italian speech . Demo uses OpenAI's [Whisper Base](https://huggingface.co/openai/whisper-base) model for speech translation, and Microsoft's
-[SpeechT5 TTS](https://huggingface.co/microsoft/speecht5_tts) model poorly fine-tuned on the VoxPopuli italian subset for text-to-speech. But it kinda sounds like Italian folks, it kinda does:
-![Cascaded STST](https://huggingface.co/datasets/huggingface-course/audio-course-images/resolve/main/s2st_cascaded.png "Diagram of cascaded speech to speech translation")
+Demo for cascaded speech-to-speech translation (STST), mapping from source speech in any language to Lithuanian. Demo uses OpenAI's [Whisper Base](https://huggingface.co/openai/whisper-base) model for speech translation, and Microsoft's
+[SpeechT5 TTS](https://huggingface.co/microsoft/speecht5_tts) 
+[Cascaded STST](https://huggingface.co/datasets/huggingface-course/audio-course-images/resolve/main/s2st_cascaded.png "Diagram of cascaded speech to speech translation")
 """
 
 demo = gr.Blocks()