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
-processor = SpeechT5Processor.from_pretrained("microsoft/speecht5_tts")
+processor = SpeechT5Processor.from_pretrained("mmhamdy/speecht5-finetuned-fleurs-it-it")
 
 model = SpeechT5ForTextToSpeech.from_pretrained("mmhamdy/speecht5-finetuned-fleurs-it-it").to(device)
 vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan").to(device)
@@ -41,7 +41,7 @@ def speech_to_speech_translation(audio):
 
 title = "Cascaded STST"
 description = """
-Demo for cascaded speech-to-speech translation (STST), mapping from source speech in English to target speech in Italian. Demo uses OpenAI's [Whisper Base](https://huggingface.co/openai/whisper-base) model for speech translation, and a finetuned Microsoft's
+Demo for cascaded speech-to-speech translation (STST), mapping from source speech in any language to target speech in Italian. Demo uses OpenAI's [Whisper Base](https://huggingface.co/openai/whisper-base) model for speech translation, and a finetuned Microsoft's
 [SpeechT5 TTS](https://huggingface.co/mmhamdy/speecht5-finetuned-fleurs-it-it) model 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")