Update app.py

app.py

@@ -16,7 +16,6 @@ device = "cuda:0" if torch.cuda.is_available() else "cpu"
 
 asr_pipe = pipeline("automatic-speech-recognition", model="openai/whisper-base", device=device, generate_kwargs = {"task": "translate"})
 
-
 m2m100_en_sw = pipeline('translation', 'facebook/m2m100_418M', src_lang='en', tgt_lang="sw")
 
 # load text-to-speech checkpoint and speaker embeddings
@@ -55,8 +54,8 @@ 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 English. 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 for text-to-speech:
+Demo for cascaded speech-to-speech translation (STST), mapping from source speech in any language to target speech in Swahili. Demo uses OpenAI's [Whisper Base](https://huggingface.co/openai/whisper-base) model for speech transcription, Meta's [M2M100](https://huggingface.co/facebook/m2m100_418M) for translation, and Microsoft's
+[SpeechT5 TTS](https://huggingface.co/samuelleecong/speecht5_finetuned_swahili) that I fine-tuned model for Swahili 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")
 """