Update app.py

app.py

@@ -1,4 +1,3 @@
-
 import gradio as gr
 import numpy as np
 import torch
@@ -12,7 +11,7 @@ device = "cuda:0" if torch.cuda.is_available() else "cpu"
 asr_pipe = pipeline("automatic-speech-recognition", model="openai/whisper-large-v2", device=device)
 
 # load text-to-speech checkpoint and speaker embeddings
-model_id = "ckandemir/speecht5_finetuned_voxpopuli_fr"  # update with your model id
+model_id = "Sandiago21/speecht5_finetuned_facebook_voxpopuli_spanish"  # update with your model id
 # pipe = pipeline("automatic-speech-recognition", model=model_id)
 model = SpeechT5ForTextToSpeech.from_pretrained(model_id)
 vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan")
@@ -22,12 +21,19 @@ speaker_embeddings = torch.tensor(embeddings_dataset[7440]["xvector"]).unsqueeze
 processor = SpeechT5Processor.from_pretrained(model_id)
 
 replacements = [
-    ("à", "a"), ("â", "a"),
-    ("ç", "c"),
-    ("é", "e"), ("è", "e"), ("ê", "e"), ("ë", "e"),
-    ("î", "i"), ("ï", "i"),
-    ("ô", "o"),
-    ("ù", "u"), ("û", "u"),
+     ("á", "a"),
+     ("ç", "c"),
+     ("è", "e"),
+     ("ì", "i"),
+     ("í", "i"),
+     ("ò", "o"),
+     ("ó", "o"),
+     ("ù", "u"),
+     ("ú", "u"),
+     ("š", "s"),
+     ("ï", "i"),
+     ("ñ", "n"),
+     ("ü", "u"),
 ]
 
 def cleanup_text(text):
@@ -43,7 +49,7 @@ def synthesize_speech(text):
     return gr.Audio.update(value=(16000, speech.cpu().numpy()))
 
 def translate(audio):
-    outputs = asr_pipe(audio, max_new_tokens=256, generate_kwargs={"task": "transcribe", "language": "french"})
+    outputs = asr_pipe(audio, max_new_tokens=256, generate_kwargs={"task": "transcribe", "language": "spanish"})
     return outputs["text"]
 
 
@@ -61,13 +67,13 @@ def speech_to_speech_translation(audio):
     return 16000, synthesised_speech
 
 
-
 title = "Cascaded STST"
 description = """
-Demo for cascaded speech-to-speech translation (STST), mapping from source speech in any language to target speech in French. Demo uses OpenAI's [Whisper Large v2](https://huggingface.co/openai/whisper-large-v2) model for speech translation, and [ckandemir/speecht5_finetuned_voxpopuli_fr](https://huggingface.co/ckandemir/speecht5_finetuned_voxpopuli_fr) checkpoint for text-to-speech, which is based on Microsoft's
-[SpeechT5 TTS](https://huggingface.co/microsoft/speecht5_tts) model for text-to-speech, fine-tuned in French Audio dataset:
+Demo for cascaded speech-to-speech translation (STST), mapping from source speech in any language to target speech in Spanish. Demo uses OpenAI's [Whisper Large v2](https://huggingface.co/openai/whisper-large-v2) model for speech translation, and [Sandiago21/speecht5_finetuned_facebook_voxpopuli_spanish](https://huggingface.co/Sandiago21/speecht5_finetuned_facebook_voxpopuli_spanish) checkpoint for text-to-speech, which is based on Microsoft's
+[SpeechT5 TTS](https://huggingface.co/microsoft/speecht5_tts) model for text-to-speech, fine-tuned in Spanish Audio dataset:
 ![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()
 
 mic_translate = gr.Interface(